Bug #39901 | Falcon crash when runing DBT2 with Stored Procedure enabled | ||
---|---|---|---|
Submitted: | 7 Oct 2008 8:53 | Modified: | 5 May 2009 13:09 |
Reporter: | Xuekun Hu | Email Updates: | |
Status: | Can't repeat | Impact on me: | |
Category: | MySQL Server: Falcon storage engine | Severity: | S2 (Serious) |
Version: | 6.0-falcon | OS: | Linux (SLES10SP1(2.6.16.46-0.12-smp)) |
Assigned to: | Hakan Küçükyılmaz | CPU Architecture: | Any |
Tags: | DBT2, F_LIMIT |
[7 Oct 2008 8:53]
Xuekun Hu
[7 Oct 2008 9:07]
Xuekun Hu
I forgot to paste the Falcon configurations. --skip_innodb \ --default_storage_engine=Falcon \ --falcon_io_threads=5 \ --falcon_record_memory_max=650M \ --falcon_page_cache_size=1650M \ --falcon_page_size=4096 \ --falcon_serial_log_buffers=100 \
[7 Oct 2008 11:55]
Hakan Küçükyılmaz
Xuekun, can you please retry with --falcon_page_size=8k and see whether you get the same assertion?
[7 Oct 2008 11:59]
Hakan Küçükyılmaz
Also, please try latest Falcon from BZR. We fixed this assertion lately Error: assertion (idx->offset == 0 || (idx->offset >= indexEnd && idx->offset + idx->length <= dbb->pageSize)) failed at line 87 in file DataPage.cpp https://code.launchpad.net/~mysql/mysql-server/mysql-6.0-falcon
[8 Oct 2008 6:10]
Xuekun Hu
Hakan, I tried the latest code. Falcon still crash. But sometimes the assertion were at diffenet place. And sometimes recovery succeeded, sometimes recovery failed. I also tried different falcon_page_size (4k, 8k, 16k), crash as same. The assertion could be [Falcon] Error: assertion (key - (UCHAR*) indexNode < 14) failed at line 109 in file IndexNode.h or [Falcon] Error: assertion (offset + length <= MAX_PHYSICAL_KEY_LENGTH) failed at line 148 in file IndexNode.h ... mysqld: my_new.cc:50: int __cxa_pure_virtual(): Assertion `"Pure virtual method called." == "Aborted"' failed. And sometimes Falcon can identify the below query cause its crash. select c_id from customer where c_w_id = 1 and c_d_id = 2 and c_last = 'ABLEABLEOUGHT' order by c_first asc limit 1 or select c_id from customer where c_w_id = 1 and c_d_id = 4 and c_last = 'PRESCALLYCALLY' order by c_first asc limit 1; And recovery failed at /home/mysql/mysql/libexec/mysqld(my_print_stacktrace+0x2e)[0x9b47ce] /home/mysql/mysql/libexec/mysqld(handle_segfault+0x33b)[0x60b16b] /lib64/libpthread.so.0[0x2b868e137e70] /lib64/libpthread.so.0(raise+0x2d)[0x2b868e137d4d] /home/mysql/mysql/libexec/mysqld(_ZN5Error5errorEPKcz+0xef)[0x82e5bf] /home/mysql/mysql/libexec/mysqld(_ZN5Cache9fetchPageEP3Dbbi8PageType8LockType+0xf5)[0x80de45] /home/mysql/mysql/libexec/mysqld(_ZN3Dbb11handoffPageEP3Bdbi8PageType8LockType+0x2a)[0x825fca] /home/mysql/mysql/libexec/mysqld(_ZN13IndexRootPage17findInsertionLeafEP3DbbiP8IndexKeyij+0xbe)[0x83b27e] /home/mysql/mysql/libexec/mysqld(_ZN13IndexRootPage13addIndexEntryEP3DbbiP8IndexKeyij+0x6f)[0x83b56f] /home/mysql/mysql/libexec/mysqld(_ZN13IndexRootPage10indexMergeEP3DbbiP14SRLUpdateIndexj+0x4dd)[0x83c61d] /home/mysql/mysql/libexec/mysqld(_ZN9SerialLog7recoverEv+0x496)[0x875546] /home/mysql/mysql/libexec/mysqld(_ZN8Database12openDatabaseEPKc+0xf6)[0x821296] /home/mysql/mysql/libexec/mysqld(_ZN10Connection11getDatabaseEPKcS1_P7Threads+0xcc)[0x815b3c] /home/mysql/mysql/libexec/mysqld(_ZN10Connection12openDatabaseEPKcS1_S1_S1_S1_P7Threads+0xa7)[0x8184a7] /home/mysql/mysql/libexec/mysqld(_ZN15StorageDatabase17getOpenConnectionEv+0x58)[0x7e2c08] /home/mysql/mysql/libexec/mysqld(_ZN14StorageHandler10initializeEv+0x99)[0x7e5d19] /home/mysql/mysql/libexec/mysqld(_ZN16StorageInterface11falcon_initEPv+0x159)[0x7ddb19] /home/mysql/mysql/libexec/mysqld(_Z24ha_initialize_handlertonP13st_plugin_int+0x38)[0x6fe198] /home/mysql/mysql/libexec/mysqld[0x7795fa] /home/mysql/mysql/libexec/mysqld(_Z11plugin_initPiPPci+0x8c9)[0x77c0e9] /home/mysql/mysql/libexec/mysqld[0x60f42f] /home/mysql/mysql/libexec/mysqld(main+0x805)[0x610535] /lib64/libc.so.6(__libc_start_main+0xf4)[0x2b868f5668b4] /home/mysql/mysql/libexec/mysqld(__gxx_personality_v0+0x271)[0x548c79] [Falcon] Error: page 131238/1 wrong page type, expected 5 got 7 Thx, Xuekun
[8 Oct 2008 14:17]
Kevin Lewis
Verified based on the ltest tests by Xuekun. Ann, one of these asserts is the same as Bug#38130.
[16 Oct 2008 22:01]
Kevin Lewis
Ann, The first crash is a duplicate of Bug#38130. The second is during recovery after that. If the recovery problem is directly due to 38130, then this can be marked as a duplicate.
[15 Dec 2008 23:41]
Kevin Lewis
Hakan, this one has been on the shelf for a while. Can you re-verify that this occurs, and if so, which call stack is the problem?
[5 May 2009 13:09]
Kevin Lewis
On Monday, May 04, 2009 10:48 PM, Kevin wrote; Xuekun, can you still reproduce Bug#39901? There have been some related fixes in the last couple months. On Tue, 05 May 2009 16:34:27 +0800, Xuekun wrote; Great. The falcon crash is gone. I tested on the latest falcon tree (rev2697, 2009-04-28) with sp enabled. And It is interesting to compare with and without SP of DBT2 performance. I found at lower concurrencies, with SP performance is lower than w/o SP performance, then with increasing the number of connections, with SP performance continue increase, then at higher concurrencies, its performance is greater than w/o sp performance. 100WH Connections 8 16 32 64 128 256 w/o sp 24792.53 28360.72 26704.51 25670.40 24302.21 24756.44 w sp 8279.95 9794.25 13844.52 20264.57 27614.40 31555.89 10WH Connections 8 16 32 64 128 256 w/o sp 41427.44 54097.19 54418.99 52233.47 49891.91 50202.14 w sp 33148.94 39251.66 44105.63 49326.23 60283.92 59840.30