Skip to content
This repository has been archived by the owner on Jul 26, 2022. It is now read-only.

Use a compaction filter to eventually remove data after DROP TABLE and DROP INDEX #7

Closed
jonahcohen opened this issue Jan 7, 2015 · 1 comment
Closed

Use a compaction filter to eventually remove data after DROP TABLE and DROP INDEX #7

jonahcohen opened this issue Jan 7, 2015 · 1 comment
Assignees
@jonahcohen

Comments

@jonahcohen
Copy link

From @mdcallag:

How do we do DROP TABLE and DROP INDEX? Using a column family per index makes this easy, just drop the CF but I assume we must support N tables or indexes per column family and need a compaction filter to eventually remove deleted rows along with an option to trigger manual compaction after dropping a large table or index to reclaim disk space faster. Assuming each key for RocksDB has the table/index ID as the prefix do we track IDs for valid objects or IDs for dropped objects that have key-value pairs that must be removed? If we track valid IDs then compaction filters must always do a bloom filter check or hash table lookup to confirm that the IDs in the merged rows are valid. And this can be a big CPU overhead. If we track IDs for dropped objects then we have a new statement management problem. This state must be persisted to survive mysqld restart. We must also figure out when an ID can be removed from this set. But the benefit is that we only need the compaction filter when that set of IDs is not empty. As an optimization we can avoid the compaction filter checks for merges between L0 and L1 because that must be as fast as possible.
@jonahcohen jonahcohen self-assigned this Jan 7, 2015
@yoshinorim yoshinorim changed the title Implement repeatable read transaction isolation for RocksDBUse a compaction filter to eventually remove data after DROP TABLE and DROP INDEX Use a compaction filter to eventually remove data after DROP TABLE and DROP INDEX Jan 14, 2015
@jonahcohen
Copy link
Author

https://reviews.facebook.net/D33069

@maykov maykov mentioned this issue Apr 29, 2015
Closed
@elenst elenst mentioned this issue May 24, 2015
Closed
maykov added a commit that referenced this issue Jul 23, 2015
@maykov
Testing the ASAN failure
eeb5ef2 
Summary:
I saw this intermittent ASAN failure
==662590== ERROR: AddressSanitizer: SEGV on unknown address 0x000000000098 (pc 0x00000112fc64 sp 0x7f0324c71710 bp 0x7f0324c71760 T21)
AddressSanitizer can not provide additional info.
    #0 0x112fc63 in my_pthread_fastmutex_lock /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/mysys/thr_mutex.c:479
    #1 0x18fb4b0 in _ZN23Dropped_indices_manager14remove_indicesERKSt13unordered_setIjSt4hashIjESt8equal_toIjESaIjEEP12Dict_manager /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/include/mysql/psi/mysql_thread.h:688
    #2 0x18b031a in _Z17drop_index_threadPv /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/storage/rocksdb/ha_rocksdb.cc:4927
    #3 0x7f0337d611e8 in _ZN6__asan10AsanThread11ThreadStartEv ??:0
    #4 0x7f03376dbfa7 in start_thread ??:0
    #5 0x7f0335a5f5bc in __clone ??:0
Thread T21 created by T0 here:
    #0 0x7f0337d5121b in pthread_create ??:0
    #1 0x18aec6a in _ZL17rocksdb_init_funcPv /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/storage/rocksdb/ha_rocksdb.cc:1765
    #2 0x60a039 in _Z24ha_initialize_handlertonP13st_plugin_int /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/handler.cc:673
    #3 0xab04e0 in _ZL17plugin_initializeP13st_plugin_int /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/sql_plugin.cc:1137
    #4 0xac8fe3 in _Z11plugin_initPiPPci /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/sql_plugin.cc:1431
    #5 0x5ef74c in _ZL22init_server_componentsv .cc:5482
    #6 0x5f0cce in _Z11mysqld_mainiPPc .cc:6254
    #7 0x7f033597defe in __libc_start_main ??:0
    #8 0x5d86a8 in _start /home/engshare/third-party2/glibc/2.17/src/glibc-2.17/csu/../sysdeps/x86_64/start.S:123
==662590== ABORTING
It looks like the Dropped_indices_manager mutex is accessed by the thread after the mutex was cleaned up. I'm moving the cleanup of the manager to the thread in hopes that this will fix the failure.

Test Plan: ran rocksdb suite, nothing failed

Reviewers: hermanlee4

Reviewed By: hermanlee4

Differential Revision: https://reviews.facebook.net/D42909
maykov added a commit that referenced this issue Aug 27, 2015
@maykov
Testing the ASAN failure
28c20ef 
Summary:
I saw this intermittent ASAN failure
==662590== ERROR: AddressSanitizer: SEGV on unknown address 0x000000000098 (pc 0x00000112fc64 sp 0x7f0324c71710 bp 0x7f0324c71760 T21)
AddressSanitizer can not provide additional info.
    #0 0x112fc63 in my_pthread_fastmutex_lock /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/mysys/thr_mutex.c:479
    #1 0x18fb4b0 in _ZN23Dropped_indices_manager14remove_indicesERKSt13unordered_setIjSt4hashIjESt8equal_toIjESaIjEEP12Dict_manager /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/include/mysql/psi/mysql_thread.h:688
    #2 0x18b031a in _Z17drop_index_threadPv /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/storage/rocksdb/ha_rocksdb.cc:4927
    #3 0x7f0337d611e8 in _ZN6__asan10AsanThread11ThreadStartEv ??:0
    #4 0x7f03376dbfa7 in start_thread ??:0
    #5 0x7f0335a5f5bc in __clone ??:0
Thread T21 created by T0 here:
    #0 0x7f0337d5121b in pthread_create ??:0
    #1 0x18aec6a in _ZL17rocksdb_init_funcPv /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/storage/rocksdb/ha_rocksdb.cc:1765
    #2 0x60a039 in _Z24ha_initialize_handlertonP13st_plugin_int /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/handler.cc:673
    #3 0xab04e0 in _ZL17plugin_initializeP13st_plugin_int /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/sql_plugin.cc:1137
    #4 0xac8fe3 in _Z11plugin_initPiPPci /data/users/jenkins/workspace/myrocks-prod/BUILD_TYPE/ASan/CLIENT_MODE/Sync/PAGE_SIZE/32/TEST_SET/MixedOther/label/mysql/mysql/sql/sql_plugin.cc:1431
    #5 0x5ef74c in _ZL22init_server_componentsv .cc:5482
    #6 0x5f0cce in _Z11mysqld_mainiPPc .cc:6254
    #7 0x7f033597defe in __libc_start_main ??:0
    #8 0x5d86a8 in _start /home/engshare/third-party2/glibc/2.17/src/glibc-2.17/csu/../sysdeps/x86_64/start.S:123
==662590== ABORTING
It looks like the Dropped_indices_manager mutex is accessed by the thread after the mutex was cleaned up. I'm moving the cleanup of the manager to the thread in hopes that this will fix the failure.

Test Plan: ran rocksdb suite, nothing failed

Reviewers: hermanlee4

Reviewed By: hermanlee4

Differential Revision: https://reviews.facebook.net/D42909
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees

@jonahcohen jonahcohen

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@jonahcohen