A site would not update

[Aussie]

I have one site on a server in France, which is a vanilla installation of CP. I visited it to try something and saw it was saying there was an update for v1.1.1

I clicked on update, but it didn’t happen. I then noticed the disk usage was around 150MB. There were multiple instances of past failed attempts to upgrade in .cagefs>tmp

In wp-content>upgrade the files in the temporary folder did not contain all that they should. I increased the script execution time in php.ini to 300 seconds, but the update stalled well before that. The screen also remained blank during that time, with no progress messages.

There was no maintenance file left over in the root directory, and it didn’t seem to be going into maintenance mode at all.

So I deleted all the temporary files to bring the disk usage back to around 35MB and uploaded a ‘no- content’ version via FTP. The site was then happy and the update notice went away.

Apart from the French location, the only difference between that server and all my others is it doesn’t run LiteSpeed, only Apache.

Just thought I’d mention it, in case it highlights a potential problem. There were no cPanel errors or error log file entries. It just seemed incapable of updating itself. It is running PHP 7.2

[james]

I clicked on update, but it didn’t happen. I then noticed the disk usage was around 150MB. There were multiple instances of past failed attempts to upgrade in .cagefs>tmp

You may have already checked this, but what was the disk quota for this site?

In wp-content>upgrade the files in the temporary folder did not contain all that they should.

This also seems relevant. What files were there and/or missing?

Another thing to check would be the memory limit for the site. However I’d expect something to show up in the error logs if that were the cause…

[Aussie]

@james
I had allowed 500MB disk space. I set 256MB for the memory limit in php.ini

I didn’t note down which files were missing, but there were very few in the root folder of the upgrade temporary folder. Of course they’re gone now.

What I do see however, is a lot of IO limit faults during that time. The limit is only 1500KB/s. My other servers do not impose strict limits.

Could this be preventing the update from completing, by interrupting the communication?

[ozfiddler]

It sounds like a similar problem to the one I had, discussed here:

Sites freezing in middle of update process

Expected behavior Update process should proceed automatically, quickly and smoothly. Current behavior On a few sites I have had to do manual upgrades. On one site the process freezes at “unpacking the update”, on two others it freezes in maintenance mode at “Copying the required files”. [freeze] I have had to restore backups to get the site live again so can’t upgrade to latest version.

Never really got to the bottom of it, but it seems to be host related. On some sites I was finding the updates folder had failed upgrade files left in there, sometimes multiple folders of them. Sometimes it froze in maintenance mode, but it often never got that far.

I’d be interested to know if it will auto-update. With the last version release all my sites auto-updated perfectly (once I’d cleaned up the wp-config files).

[james]

What I do see however, is a lot of IO limit faults during that time. The limit is only 1500KB/s. My other servers do not impose strict limits.

Could this be preventing the update from completing, by interrupting the communication?

It depends on what happens when there is an IO limit fault. If the corresponding IO operation fails (is blocked/aborted) then yes, that could definitely cause what you’re seeing.

If possible it would be better to throttle operations that go over the limit (make the offending read or write take a few extra milliseconds instead of just failing completely).

The WP/CP update process is not designed to deal with servers that may fail when writing or reading files. It could be made to deal with these conditions but that would be a pretty major undertaking.

[Aussie]

I knew I’d seen a similar thread on this, but couldn’t find it by looking at the titles last night. Brain freeze.

This server is a budget one and I only used it because it is a master reseller account. It’s very slow, especially from here in Australia.

It also refused to install the CP Brute Force Login Protection plugin and I had to use the WP version.

I would get rid of it, if it wasn’t for the MR capability. I am pretty sure the IO limiting is the culprit now. It makes perfect sense.

Thanks for the help and link to the other thread.

[Aussie]

I’d be interested to know if it will auto-update.

If you’re referring to my site, then no it didn’t auto-update. That’s where I saw the failed attempts that were clogging up the disk space.

[ozfiddler]

Have you got auto-update disabled anywhere? I found I had some sites with that instruction in wp-config, and I didn’t know it was there.

[Aussie]

That was one of the things I checked first. Was there anything added to the config file? No there isn’t.

I have raised a support ticket with the hosting company asking them to increase my IO limit. They will probably not answer until tomorrow, and I can bet I know what they will say

[Aussie]

Update on this issue.

The hosting company finally got back to me after a long weekend in their country.

They agreed to increase the IO limit on my entire MR reseller account from 1500KB/s to 2000KB/s.

I then deleted the files and database records for the current v1.1.1 installation and did a clean install of v1.1.0

This time the update to v1.1.1 succeeded. It caused an 1100KB/s IO spike, but proceeded as expected.

I had not been aware that they also have a 20% trigger point for IO errors. It is in their TOS. So the update did trigger another fault, but went ahead anyway.

So, the issue seems to be, that you need at least 2MB/s IO limit in your hosting for CP. Otherwise, you are going to be unable to do core updates.

[ozfiddler]

Thanks Graham. Good work. I’ll be contacting Synergy to ask if they have a similar limit.

[Aussie]

This server is the only one that has the Concurrent Usage plugin in cPanel and I’ve been looking in detail at what it’s telling me.

The IO graph says it includes database usage. For a lot of the daily faults, the traffic in KB/s is very low and should not have triggered a fault even with the 20% restriction. While at other times, the traffic was high, but did not trigger a fault.

So, I can only assume it was database usage that did it. To me, as a layman, it seems interesting that the two should be grouped together, and may help to understand why the update stalled.

[james]

that you need at least 2MB/s IO limit in your hosting for CP

Since this will depend on a number of other factors such as the server’s download speed and other simultaneous activity on the same site (number of visitors) I don’t think it’s possible to put a hard limit on this.

I would recommend not using a host that may break your site’s file I/O operations without warning.

[Aussie]

I would recommend not using a host that may break your site’s file I/O operations without warning.

I remember that it was not that long ago that even expensive reseller accounts here in Australia charged extra to upgrade to SSD hosting, so as to get I/O limits of 2048KB/s.

Admittedly, technology progresses quickly, but I think there are a lot of hosting companies that run a tight ship when it comes to reseller accounts. As always, you get what you pay for - if you’re lucky.