mysql error 2006

[mitchell-joblin]

I'm trying to analyze llvm using the provmity based analysis without specifying a revisions range (i.e. analyze entire history). The analysis hangs at one point (several hours into the analysis) and the log indicates codeface is trying indefinitely to re-establish the connection to the database.

The log shows the following events just before the error is produced: 2017-10-28 02:54:53 [codeface.util] Process-7 DEBUG: Running command: git --git-dir=/home/vagrant/git/llvm/.git show --format=full --shortstat --numstat --ignore-space-change 346818f5140094eca35697fb101c96a2d902cddf 17-10-28 02:56:01 [codeface.cluster.cluster] Process-7 DEVINFO: Shelving the VCS object 2017-10-28 02:56:44 [codeface.dbmanager] Process-7 DEVINFO: Using project llvm/proximity with ID 29 2017-10-28 02:56:44 [codeface.cluster.cluster] Process-7 DEVINFO: Reading from data base /home/vagrant/workspace/codeface_results/llvm/proximity/062--814997-562b26/vcs_analysis.db... 2017-10-28 02:57:13 [codeface.VCS] Process-7 DEVINFO: Using cached data to extract commit information 2017-10-28 02:57:13 [codeface.dbmanager] Process-7 DEVINFO: Using project llvm/proximity with ID 29 2017-10-28 03:10:52 [codeface.dbmanager] Process-7 INFO: DBE args: (2006, 'MySQL server has gone away') 2017-10-28 03:10:52 [codeface.dbmanager] Process-7 WARNING: MySQL Server gone away, trying to reconnect. 2017-10-28 03:10:55 [codeface.dbmanager] Process-7 INFO: DBE args: (2006, 'MySQL server has gone away') 2017-10-28 03:10:55 [codeface.dbmanager] Process-7 WARNING: MySQL Server gone away, trying to reconnect.

What follows is just repeating the error and the trying to reconnect message.

Any ideas?

[mitchell-joblin]

After some further look into it, it looks like the error is produced sometime during the time series analysis phases. In project.py there is the following:

#########

Global stage 1: Time series generation

log.info("=> Preparing time series data")
#dispatch_ts_analysis(project_resdir, conf)

#########
# Global stage 2: Complexity analysis
## NOTE: We rely on proper timestamps, so we can only run
## after time series generation
log.info("=> Performing complexity analysis")

I think the error are in theses phases. Im not so experienced with whats happening there, so some points would be great!

Thanks

[mitchell-joblin]

Found the problem. The issue was that the max buffer size for mysql was set too low. Iv double it now to 512MB and no longer get the error. The fix is sitting on my branch.

[wolfgangmauerer]

Am 29/10/2017 um 21:54 schrieb mitchell-joblin:

Found the problem. The issue was that the max buffer size for mysql was set too low. Iv double it now to 512MB and no longer get the error. The fix is sitting on my branch.

thanks for reporting and debugging the problem! On your branch mitchell_changes (337afd6..fd5b819), I see four commits, but none of them seems to address the change you described. Which commit are you referring to exactly? Or am I looking at the wrong branch?

Since I recall that we've had this problem in the past: Can you add an info message to the debug output that maybe increasing the buffer size will help? Since the DB's error message is (at least for a non-mysql expert like me) only vaguely connected to the root cause, we might want to remind ourselves when this happens next time in three years or so.

Thanks, Wolfgang

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/siemens/codeface/issues/63#issuecomment-340293710, or mute the thread https://github.com/notifications/unsubscribe-auth/ABbE14YHGc231qWdeHfCzTrENksmkazlks5sxOXxgaJpZM4QJ5nH.

[mitchell-joblin]

My mistake. I pushed to the wrong branch. I've corrected that and you should now see it on the mitchell_changes branch I made the pull request with.

Good idea regarding the extra log message. I've also added that now!

Cheers,

Mitchell