[MDEV-2720] LP:632959 - Aria recovery failure "UNDO records skipped in UNDO phase; some tables may be left inconsistent" Created: 2010-09-08  Updated: 2012-10-04  Resolved: 2012-10-04

Status: Closed
Project: MariaDB Server
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Philip Stoev (Inactive) Assignee: Michael Widenius
Resolution: Not a Bug Votes: 0
Labels: Launchpad

Attachments: XML File LPexportBug632959.xml     Zip Archive LPexportBug632959_var-bug632959.zip    

 Description   

When recovering from various RQG workloads, maria_read_log reported the following:

The transaction log starts from lsn (3,0x2926)
TRACE of the last maria_read_log
/home/buildbot/rqg-buildbot-slave/rqg-weekend/build/mysql-test//../storage/maria/maria_read_log: Maria engine: starting recovery
recovered pages: 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% (1.3 seconds); transactions to roll back: 1 0**WARNING: 11776 UNDO records skipped in UNDO phase; some tables may be left inconsistent!**
(0.3 seconds);
/home/buildbot/rqg-buildbot-slave/rqg-weekend/build/mysql-test//../storage/maria/maria_read_log: Maria engine: recovery done
maria_read_log: DOUBTFUL (1 warnings, check previous output)

Apart from the underlying issue, maria_read_log returned 0, whereas instead this should have been a clear failure and the script should have returned a non-zero failure code.

RQG command line:

  1. 2010-09-07T14:40:31 # /home/buildbot/randgen/runall.pl \
  2. 2010-09-07T14:40:31 # --queries=1M \
  3. 2010-09-07T14:40:31 # --engine=Maria \
  4. 2010-09-07T14:40:31 # -mysqld=-log-output=file \
  5. 2010-09-07T14:40:31 # --reporters=ErrorLog,Backtrace,Recovery,Shutdown \
  6. 2010-09-07T14:40:31 # --duration=240 \
  7. 2010-09-07T14:40:31 # --threads=5 \
  8. 2010-09-07T14:40:31 # --rows=10000 \
  9. 2010-09-07T14:40:31 # --mask-level=1 \
  10. 2010-09-07T14:40:31 # -mysqld=-loose-maria-group-commit=hard \
  11. 2010-09-07T14:40:31 # -mysqld=-loose-maria_group_commit_interval=1 \
  12. 2010-09-07T14:40:31 # -mysqld=-maria-checkpoint-interval=1 \
  13. 2010-09-07T14:40:31 # -mysqld=-maria-block-size=16K \
  14. 2010-09-07T14:40:31 # --grammar=conf/engines/varchar.yy \
  15. 2010-09-07T14:40:31 # --gendata=conf/engines/varchar.zz \
  16. 2010-09-07T14:40:31 # --mask=25153 \
  17. 2010-09-07T14:40:31 # --queries=100000000 \
  18. 2010-09-07T14:40:31 # --mask=25153 \
  19. 2010-09-07T14:40:31 # --basedir=. \
  20. 2010-09-07T14:40:31 # --seed=time \
  21. 2010-09-07T14:40:31 # --vardir=/home/buildbot/rqg-buildbot-slave/rqg-weekend/vardir-maria_recovery_weekend/current

bzr version-info:

revision-id: <email address hidden>
date: 2010-09-06 02:48:51 +0300
build-date: 2010-09-08 10:21:55 +0300
revno: 2922
branch-nick: 5.1

 Comments   
Comment by Philip Stoev (Inactive) [ 2010-09-08 ]

Re: Aria recovery failure "UNDO records skipped in UNDO phase; some tables may be left inconsistent"

Comment by Philip Stoev (Inactive) [ 2010-09-08 ]

datadir, logs, etc.
LPexportBug632959_var-bug632959.zip

Comment by Philip Stoev (Inactive) [ 2010-09-08 ]

Re: Aria recovery failure "UNDO records skipped in UNDO phase; some tables may be left inconsistent"
This may be caused by the fact that only the 3rd part of the log was available to maria_read_log. The first 2 parts were rotated out.

Comment by Rasmus Johansson (Inactive) [ 2010-09-13 ]

Launchpad bug id: 632959

Generated at Thu Feb 08 06:43:45 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.