rllmukforum Posted October 21, 2016 Posted October 21, 2016 Last night I began manually uploading the delta upgrade file from 4.1.15 to 4.1.16.1. This was about 3am when I board load was very light. I left the board open and continued posting as I uploaded the files, as I have for years of upgrades prior to this. Half way through uploading the files, I noticed the site wouldn't respond. I then noticed the sql partition was full, and eventually tracked it down to the core_log.idb table ballooning to 22gig. As fast as I could delete files on the partition, this table continued to grow. To cut a 3 hour triage story short, I had to restore the http root files, drop the database and import a backup taken prior to the upgrade. Is there any obvious reason as to why this might have happened? I've always left the board open while uploading, is this not best practice? I assumed it wouldn't cause a conflict as empirically it hasn't and the upgrade script itself doesn't take the forum offline until it starts to execute changes. For what it's worth, with no other actions other then restarting nginx and php, the board came straight back to life after the db was restored. I have not tried 4.1.16.1 again, wanted to get feedback first.
MADMAN32395 Posted October 21, 2016 Posted October 21, 2016 During update the front and backend enter maintenance mode. So it's not like someone would have posted or something. It's always best practice to backup before updating any service. Also if you have more control of the software could update webserver and SQL to a newer stable release. I've seen some crazy stuff disappear after updating.
rllmukforum Posted October 23, 2016 Author Posted October 23, 2016 On 10/20/2016 at 10:49 PM, MADMAN32395 said: During update the front and backend enter maintenance mode. So it's not like someone would have posted or something. That's not when it broke, though. It was while I was uploading the files, almost as though some conflict between an old and new version caused some runaway nonsense in the database. I didn't get as far as triggering the upgrade script, which of course puts the board into maintenance mode. On 10/20/2016 at 10:49 PM, MADMAN32395 said: It's always best practice to backup before updating any service. Also if you have more control of the software could update webserver and SQL to a newer stable release. I've seen some crazy stuff disappear after updating. Yeah, I only do these updates after the nightly runs, and take a root http backup before doing so. However, in this case, MariaDB was just fine after I restored the db from a backup, so I don't believe any server configuration was at fault. Curious that no one else encountered this, judging by the lack of replies. I assume it was a freak occurrence, rather than something systemic. I'll try again and ensure the board is manually offline while I upload the files. Thanks for the feedback.
steve00 Posted October 23, 2016 Posted October 23, 2016 Have had similar experience when upgrading via admincp if I miss one upgrade version (e.g. 4.1.14.2 to 4.1.15 .... missing out the 4.1.14.3 upgrade) Had to overcome that by uploading files manually via ftp to stop the errors given by admincp upgrade, but appears your manual upgrade caused problems just by uploading the files Must admit though strange upgrading manually works for me when missing out an upgrade version or two but doesn't work when upgrading via admincp
Recommended Posts
Archived
This topic is now archived and is closed to further replies.