Can't install mods after upgrade from 2.0.19 to 2.1.4

Started by hustreamload, July 21, 2023, 11:53:29 AM

Previous topic - Next topic

shawnb61

#80
Quote from: hustreamload on August 13, 2023, 08:51:47 AMAnswers/questions on your questions:
 - Which backup options were selected? Where should I select this option?
 - Does disabling backups fix the issue? Where should I select this option?
 - What is the value of MySQL's wait_timeout? se my question above
Thanks!

The backup options are in the admin panel, under Admin | Package Manager | Options.  There are two 'Create a backup of...' options at the bottom of the form.  What are they set to?  Does it work if you disable both?

The wait_timeout is set by your host.  To view it, in the myphpadmin SQL tab, run this query:
SHOW VARIABLES LIKE 'wait_timeout';
40MB (compressed) backups are actually pretty big, for just a copy of /Sources & /Themes...  Mine are all in the 4-7MB (compressed) range.  (The entire SMF install is only ~3.5MB...)  It is possible that something in there is much bigger than expected.  I would search both /Sources & /Themes for the biggest subfolders for clues. 
Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

hustreamload

shawnb61, disabling the backup option, THAT was the solution!
Thanks a lot, I could install every mod in seconds!
MAGIC!!!!
Thank you very much!

shawnb61

Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

hustreamload

wait_timeout is 20
but without any change in it the mods are already available
thanks!

shawnb61

That wait_timeout is extremely low...

The mysql default for wait_timeout is 28800.

So what is happening here is that the backup, which is unusually large, is taking longer than 20 seconds to create, causing mysql to timeout, in part due to an unusually low timeout setting.
Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

@rjen

That explains why a 'hooks only' mod does install fine: no files needed in the backup
Running SMF 2.1 with latest TinyPortal at www.fjr-club.nl

hustreamload

Ok, maybe that explains the unavailability of "Find and repair any errors" as well, it runs still on 503 timeout. How to set it, and what is the offered value?

shawnb61

Quote from: hustreamload on August 15, 2023, 11:51:37 AMOk, maybe that explains the unavailability of "Find and repair any errors" as well, it runs still on 503 timeout. How to set it, and what is the offered value?

Your host should increase this.  20 prohibits a lot of maintenance activity.  SMF does chunk up big maintenance tasks, but as you have learned, it is pretty easy to pass 20.

If they're not OK with restoring the default of 28800, see if you can at least get them to go to a few hundred.  As big as they're willing to go.  Enough so you can run the Find and Repair any Errors.
Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

hustreamload

We are on a shared server (Hosting24), I cannot reach them to rise this level. "That can happen if you use heavy or very long MySQL queries. To solve it, try adding the mysql_reconnect command before each query. " (FAQ)
Its not clear for me, if that 20 for wait_timeot in 2.0.19 was enough, why is not enough in 2.1.4....si any idea how to solve it?

shawnb61

Quote from: hustreamload on August 21, 2023, 03:25:37 AMIts not clear for me, if that 20 for wait_timeot in 2.0.19 was enough, why is not enough in 2.1.4....si any idea how to solve it?

The logic behaves pretty much the same in 2.1 as in 2.0.  20 seconds would cause issues in 2.0, depending on the request.  You can search the forum for wait_timeout or "Mysql server has gone away" and see prior similar issues. 

Regarding your thread here - the backup option to make a backup of /Sources & /Themes didn't exist in 2.0.  That's why it didn't bother you in 2.0.  Also, we never found out why your backups are so big.

Regarding Find & Repair Errors - unfortunately, we cannot know WHY F&RE is failing for you in this instance, because F&RE is the very tool that runs such checks...

SMF needs some mechanism to run long-running tasks, e.g., Scheduled Tasks & F&RE.

If you stay with that host, the best advice I can provide is to download a copy of your DB & run F&RE locally.  Another option is to follow their recommendation to issue a reconnect before every query in F&RE.  I am not sure how extensive those changes would be.  Or you could attempt to issue a "SET session wait_timeout=1000;" somewhere in F&RE.  That would only work if they allow such an update (& probably not...).  For those last two recommendations, you're on your own & coding around an unreasonable setting.
Address the process rather than the outcome.  Then, the outcome becomes more likely.   - Fripp

Advertisement: