Server crash

[Fleiding]

Our server gets a lot of process requests from the following URL:

request URI: /forum/index.php?routestring=chat/loadheaderdata

Because of these requests it crashes and it must be restarted again. Is this a default feature in vB5 and if so, how can I disable this temporarily? I want to check if this is causing the server to crash.

[Fleiding]

Gentle bump.

[glennrocksvb]

It looks like the requests for vB Messenger chat updates. It polls for new updates every 60 seconds by default. You can make it longer or turn off chat entirely.

[Fleiding]

Thanks Glenn, I increased this number and will check if this will keep the server steady.

[Fleiding]

I've increased the settings, but I'm still facing the same issue unfortunately. Below is the data at the moment of the crash:

https://pastebin.com/raw/wFs8Mnea

[Wayne Luke]

Then you will need to turn off vBulletin Messenger. The calls are light and shouldn't cause a server to crash. However, if you have an older server or limited resources, you can be hitting that. Within the software, the only alternative to increasing the polling is to turn off the vBulletin Messenger functionality.

[Fleiding]

I've adjusted the numbers again and if a crash should occur again I'll disable vB Messenger.

[glennrocksvb]

Are you on Shared or VPS?

[Fleiding]

Vps.

[Fleiding]

We now have the same issue occuring with several requests and nothing looks out of the ordinary. I have asked my host to switch to mod_php to see if this will fix the issue.

[Fleiding]

I have disabled vB Messenger and switched to Apache, but unfortunate I still have the same issue. My host informed me that there is nothing out of the ordinary when the httpd memory reaches over 1.5 GB, so I'm not sure what else can cause this issue. After a restart of httpd everything works fine.

Would switching to nginx would solve this issue?

[Mark.B]

I think you're chasing rainbows talking of further OS switches.

You need to get your hosts involved and get them to dig down and find where the bottleneck is.

[Fleiding]

I will push them again, but they spent so much hours looking at all logs and files I provided I don't think it will help much. I'm using a VPS of one of the best Dutch providers, so it's bugging me that they can't figure out the issue yet.

In the mean time, I have set Minutes to Cache Content for Guests to 15 and Search Result Cache Lifetime to 5 to see if this helps. Does the System Cache clear automatically per standard or do I have to create a Scheduled Task? If so, how can I do this?

[Wayne Luke]

What are the sizes of your cache, cacheevent, and searchlog tables? Please provide the values for both the data and index lengths.

What is the number of rows in your searchlog table?

You can find this information in the vBulletin AdminCP under Maintenance -> Diagnostics. At the bottom is a System Information form. Set it to Table Status and submit the form. Find the above tables in the list. Once you have recorded the values, you can run these queries:

truncate table cache;
truncate table cacheevent;
truncate table searchlog;

After you run those queries, does the memory usage return to normal?

Your issue may be caused by the following issue, which currently scheduled to be released in 5.3.4: http://tracker.vbulletin.com/browse/VBV-17655

[Fleiding]

Thanks for your reply, Wayne.

Date and Index Lengths:

cache: 16384 & 16384
cacheevent: 6602752 & 4653056
searchlog: 170934272 & 11599872

After a truncate of these tables the memory usage returns to normal.