作者: 谢瑞阳 10225101483 徐翔宇 10225101535
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

226 lines
5.8 KiB

fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
fix problems in LevelDB's caching code Background: LevelDB uses a cache (util/cache.h, util/cache.cc) of (key,value) pairs for two purposes: - a cache of (table, file handle) pairs - a cache of blocks The cache places the (key,value) pairs in a reference-counted wrapper. When it returns a value, it returns a reference to this wrapper. When the client has finished using the reference and its enclosed (key,value), it calls Release() to decrement the reference count. Each (key,value) pair has an associated resource usage. The cache maintains the sum of the usages of the elements it holds, and removes values as needed to keep the sum below a capacity threshold. It maintains an LRU list so that it will remove the least-recently used elements first. The max_open_files option to LevelDB sets the size of the cache of (table, file handle) pairs. The option is not used in any other way. The observed behaviour: If LevelDB at any time used more file handles concurrently than the cache size set via max_open_files, it attempted to reduce the number by evicting entries from the table cache. This could happen most easily during compaction, and if max_open_files was low. Because the handles were in use, their reference count did not drop to zero, and so the usage sum in the cache was not modified by the evictions. Subsequent Insert() calls returned valid handles, but their entries were immediately evicted from the cache, which though empty still acted as though full. As a result, there was effectively no caching, and the number of open file handles rose []ly until it hit system-imposed limits and the process died. If one set max_open_files lower, the cache was more likely to exhibit this beahviour, and cause the process to run out of file descriptors. That is, max_open_files acted in almost exactly the opposite manner from what was intended. The problems: 1. The cache kept all elements on its LRU list eligible for capacity eviction---even those with outstanding references from clients. This was ineffective in reducing resource consumption because there was an outstanding reference, guaranteeing that the items remained. A secondary issue was that there is no guarantee that these in-use items will be the last things reached in the LRU chain, which actually recorded "least-recently requested" rather than "least-recently used". 2. The sum of usages was decremented not when a (key,value) was evicted from the cache, but when its reference count went to zero. Thus, when things were removed from the cache, either by garbage collection or via Erase(), the usage sum was not necessarily decreased. This allowed the cache to act as though full when it was in fact not, reducing caching effectiveness, and leading to more resources being consumed---the opposite of what the evictions were intended to achieve. 3. (minor) The cache's clients insert items into it by first looking up the key, and inserting only if no value is found. Although the cache has an internal lock, the clients use no locking to ensure atomicity of the Lookup/Insert pair. (see table/table.cc: block_cache->Insert() and db/table_cache.cc: cache_->Insert()). Thus, if two threads Insert() at about the same time, they can both Lookup(), find nothing, and both Insert(). The second Insert() would evict the first value, leaving each thread with a handle on its own version of the data, and with the second version in the cache. It would be better if both threads ended up with a handle on the same (key,value) pair, which implies it must be the first item inserted. This suggests that Insert() should not replace an existing value. This can be made safe with current usage inside LeveDB itself, but this is not easy to change first because Cache is a public interface, so to change the semantics of an existing call might break things, second because Cache is an abstract virtual class, so adding a new abstract virtual method may break other implementations, and third, the new method "insert without replacing" cannot be implemented in terms of the existing methods, so cannot be implemented with a non-abstract default. But fortunately, the effects of this issue are minor, so this issue is not fixed by this change. The changes: The assumption in the fixes is that it is always better to cache entries unless removal from the cache would lead to deallocation. Cache entries now have an "in_cache" boolean indicating whether the cache has a reference on the entry. The only ways that this can become false without the entry being passed to its "deleter" are via Erase(), via Insert() when an element with a duplicate key is inserted, or on destruction of the cache. The cache now keeps two linked lists instead of one. All items in the cache are in one list or the other, and never both. Items still referenced by clients but erased from the cache are in neither list. The lists are: - in-use: contains the items currently referenced by clients, in no particular order. (This list is used for invariant checking. If we removed the check, elements that would otherwise be on this list could be left as disconnected singleton lists.) - LRU: contains the items not currently referenced by clients, in LRU order A new internal Ref() method increments the reference count. If incrementing from 1 to 2 for an item in the cache, it is moved from the LRU list to the in-use list. The Unref() call now moves things from the in-use list to the LRU list if the reference count falls to 1, and the item is in the cache. It no longer adjusts the usage sum. The usage sum now reflects only what is in the cache, rather than including still-referenced items that have been evicted. The LRU_Append() now takes a "list" parameter so that it can be used to append either to the LRU list or the in-use list. Lookup() is modified to use the new Ref() call, rather than adjusting the reference count and LRU chain directly. Insert() eviction code is also modified to adjust the usage sum and the in_cache boolean of the evicted elements. Some LevelDB tests assume that there will be no caching whatsoever if the cache size is set to zero, so this is handled as a special case. A new private method FinishErase() is factored out with the common code from where items are removed from the cache. Erase() is modified to adjust the usage sum and the in_cache boolean of the erased elements, and to use FinishErase(). Prune() is modified to use FinishErase() also, and to make use of the fact that the lru_ list now contains only items with reference count 1. - EvictionPolicy is modified to test that an entry with an outstanding handle is not evicted. This test fails with the old cache.cc. - A new test case UseExceedsCacheSize verifies that even when the cache is overfull of entries with outstanding handles, none are evicted. This test fails with the old cache.cc, and is the key issue that causes file descriptors to run out when the cache size is set too small. ------------- Created by MOE: https://github.com/google/moe MOE_MIGRATED_REVID=123247237
8 years ago
  1. // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
  2. // Use of this source code is governed by a BSD-style license that can be
  3. // found in the LICENSE file. See the AUTHORS file for names of contributors.
  4. #include "leveldb/cache.h"
  5. #include <vector>
  6. #include "util/coding.h"
  7. #include "util/testharness.h"
  8. namespace leveldb {
  9. // Conversions between numeric keys/values and the types expected by Cache.
  10. static std::string EncodeKey(int k) {
  11. std::string result;
  12. PutFixed32(&result, k);
  13. return result;
  14. }
  15. static int DecodeKey(const Slice& k) {
  16. assert(k.size() == 4);
  17. return DecodeFixed32(k.data());
  18. }
  19. static void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); }
  20. static int DecodeValue(void* v) { return reinterpret_cast<uintptr_t>(v); }
  21. class CacheTest {
  22. public:
  23. static CacheTest* current_;
  24. static void Deleter(const Slice& key, void* v) {
  25. current_->deleted_keys_.push_back(DecodeKey(key));
  26. current_->deleted_values_.push_back(DecodeValue(v));
  27. }
  28. static const int kCacheSize = 1000;
  29. std::vector<int> deleted_keys_;
  30. std::vector<int> deleted_values_;
  31. Cache* cache_;
  32. CacheTest() : cache_(NewLRUCache(kCacheSize)) { current_ = this; }
  33. ~CacheTest() { delete cache_; }
  34. int Lookup(int key) {
  35. Cache::Handle* handle = cache_->Lookup(EncodeKey(key));
  36. const int r = (handle == nullptr) ? -1 : DecodeValue(cache_->Value(handle));
  37. if (handle != nullptr) {
  38. cache_->Release(handle);
  39. }
  40. return r;
  41. }
  42. void Insert(int key, int value, int charge = 1) {
  43. cache_->Release(cache_->Insert(EncodeKey(key), EncodeValue(value), charge,
  44. &CacheTest::Deleter));
  45. }
  46. Cache::Handle* InsertAndReturnHandle(int key, int value, int charge = 1) {
  47. return cache_->Insert(EncodeKey(key), EncodeValue(value), charge,
  48. &CacheTest::Deleter);
  49. }
  50. void Erase(int key) { cache_->Erase(EncodeKey(key)); }
  51. };
  52. CacheTest* CacheTest::current_;
  53. TEST(CacheTest, HitAndMiss) {
  54. ASSERT_EQ(-1, Lookup(100));
  55. Insert(100, 101);
  56. ASSERT_EQ(101, Lookup(100));
  57. ASSERT_EQ(-1, Lookup(200));
  58. ASSERT_EQ(-1, Lookup(300));
  59. Insert(200, 201);
  60. ASSERT_EQ(101, Lookup(100));
  61. ASSERT_EQ(201, Lookup(200));
  62. ASSERT_EQ(-1, Lookup(300));
  63. Insert(100, 102);
  64. ASSERT_EQ(102, Lookup(100));
  65. ASSERT_EQ(201, Lookup(200));
  66. ASSERT_EQ(-1, Lookup(300));
  67. ASSERT_EQ(1, deleted_keys_.size());
  68. ASSERT_EQ(100, deleted_keys_[0]);
  69. ASSERT_EQ(101, deleted_values_[0]);
  70. }
  71. TEST(CacheTest, Erase) {
  72. Erase(200);
  73. ASSERT_EQ(0, deleted_keys_.size());
  74. Insert(100, 101);
  75. Insert(200, 201);
  76. Erase(100);
  77. ASSERT_EQ(-1, Lookup(100));
  78. ASSERT_EQ(201, Lookup(200));
  79. ASSERT_EQ(1, deleted_keys_.size());
  80. ASSERT_EQ(100, deleted_keys_[0]);
  81. ASSERT_EQ(101, deleted_values_[0]);
  82. Erase(100);
  83. ASSERT_EQ(-1, Lookup(100));
  84. ASSERT_EQ(201, Lookup(200));
  85. ASSERT_EQ(1, deleted_keys_.size());
  86. }
  87. TEST(CacheTest, EntriesArePinned) {
  88. Insert(100, 101);
  89. Cache::Handle* h1 = cache_->Lookup(EncodeKey(100));
  90. ASSERT_EQ(101, DecodeValue(cache_->Value(h1)));
  91. Insert(100, 102);
  92. Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
  93. ASSERT_EQ(102, DecodeValue(cache_->Value(h2)));
  94. ASSERT_EQ(0, deleted_keys_.size());
  95. cache_->Release(h1);
  96. ASSERT_EQ(1, deleted_keys_.size());
  97. ASSERT_EQ(100, deleted_keys_[0]);
  98. ASSERT_EQ(101, deleted_values_[0]);
  99. Erase(100);
  100. ASSERT_EQ(-1, Lookup(100));
  101. ASSERT_EQ(1, deleted_keys_.size());
  102. cache_->Release(h2);
  103. ASSERT_EQ(2, deleted_keys_.size());
  104. ASSERT_EQ(100, deleted_keys_[1]);
  105. ASSERT_EQ(102, deleted_values_[1]);
  106. }
  107. TEST(CacheTest, EvictionPolicy) {
  108. Insert(100, 101);
  109. Insert(200, 201);
  110. Insert(300, 301);
  111. Cache::Handle* h = cache_->Lookup(EncodeKey(300));
  112. // Frequently used entry must be kept around,
  113. // as must things that are still in use.
  114. for (int i = 0; i < kCacheSize + 100; i++) {
  115. Insert(1000 + i, 2000 + i);
  116. ASSERT_EQ(2000 + i, Lookup(1000 + i));
  117. ASSERT_EQ(101, Lookup(100));
  118. }
  119. ASSERT_EQ(101, Lookup(100));
  120. ASSERT_EQ(-1, Lookup(200));
  121. ASSERT_EQ(301, Lookup(300));
  122. cache_->Release(h);
  123. }
  124. TEST(CacheTest, UseExceedsCacheSize) {
  125. // Overfill the cache, keeping handles on all inserted entries.
  126. std::vector<Cache::Handle*> h;
  127. for (int i = 0; i < kCacheSize + 100; i++) {
  128. h.push_back(InsertAndReturnHandle(1000 + i, 2000 + i));
  129. }
  130. // Check that all the entries can be found in the cache.
  131. for (int i = 0; i < h.size(); i++) {
  132. ASSERT_EQ(2000 + i, Lookup(1000 + i));
  133. }
  134. for (int i = 0; i < h.size(); i++) {
  135. cache_->Release(h[i]);
  136. }
  137. }
  138. TEST(CacheTest, HeavyEntries) {
  139. // Add a bunch of light and heavy entries and then count the combined
  140. // size of items still in the cache, which must be approximately the
  141. // same as the total capacity.
  142. const int kLight = 1;
  143. const int kHeavy = 10;
  144. int added = 0;
  145. int index = 0;
  146. while (added < 2 * kCacheSize) {
  147. const int weight = (index & 1) ? kLight : kHeavy;
  148. Insert(index, 1000 + index, weight);
  149. added += weight;
  150. index++;
  151. }
  152. int cached_weight = 0;
  153. for (int i = 0; i < index; i++) {
  154. const int weight = (i & 1 ? kLight : kHeavy);
  155. int r = Lookup(i);
  156. if (r >= 0) {
  157. cached_weight += weight;
  158. ASSERT_EQ(1000 + i, r);
  159. }
  160. }
  161. ASSERT_LE(cached_weight, kCacheSize + kCacheSize / 10);
  162. }
  163. TEST(CacheTest, NewId) {
  164. uint64_t a = cache_->NewId();
  165. uint64_t b = cache_->NewId();
  166. ASSERT_NE(a, b);
  167. }
  168. TEST(CacheTest, Prune) {
  169. Insert(1, 100);
  170. Insert(2, 200);
  171. Cache::Handle* handle = cache_->Lookup(EncodeKey(1));
  172. ASSERT_TRUE(handle);
  173. cache_->Prune();
  174. cache_->Release(handle);
  175. ASSERT_EQ(100, Lookup(1));
  176. ASSERT_EQ(-1, Lookup(2));
  177. }
  178. TEST(CacheTest, ZeroSizeCache) {
  179. delete cache_;
  180. cache_ = NewLRUCache(0);
  181. Insert(1, 100);
  182. ASSERT_EQ(-1, Lookup(1));
  183. }
  184. } // namespace leveldb
  185. int main(int argc, char** argv) { return leveldb::test::RunAllTests(); }