Topic: need a little advice on tuning Apache 2.4

[Megunticook]

I manage a WordPress site on an AWS Linux instance running Apache 2.4. Just had an issue today where the server became unresponsive, so I rebooted and it came back. Then I looked at the log and it appears Apache ran out of memory, although the AWS metrics indicate that the server never used more than 33% of available memory (4GB).

I've pasted some of the httpd_error_log file in below from the point where the "out of memory" warnings began. But from my research online I'm thinking I need to specify ServerLimit and MaxRequestWorkers values in my httpd.conf file. Right now there is nothing specified at all there, so I'm thinking Apache must be using default values.

Is this a good place to start? Based on a blog post I read ( https://aws.amazon.com/premiumsupport/knowledge-center/ec2-apache-memory-tuning/ ), I'm thinking I would add this to my httpd.conf file. Please tell me if I'm on the right track or should look at other options:

[James Blond]

The first error PHP Fatal error: Out of memory is PHP related, not apache. See memory_limit in your php.ini

That HTTP/2 is not working is because of prefork mpm. You would have to use worker or event mpm to use that. However than you can use PHP as module but must run it over mod_fcgid or FPM.

[Megunticook]

Thanks for your reply.

PHP memory limit is set at 128M in php.ini. I changed it to 256M.

Couldn't those PHP errors be a result of Apache hogging all the available memory? Or is the log telling me that the source of the problem is PHP?

I think if WordPress was running out of memory I would get an error message like this:

[Megunticook]

I'm monitoring the new server, which is now using the event MPM instead of prefork.

After running the top command, it's telling me that it's using 3.2G our of 4G total memory, which seems awfully high given that the site is only seeing light to moderate use. It says there are 133 total tasks, 1 running, 90 sleeping.

Do I need to make some adjustments here? Seems odd that it would be consuming 80% of available memory when there's such a light load on the server.

What are all those sleeping procecess? Are they consuming a bunch of memory? Can I reduce them? Something doesn't seem right here...

I just added this to my httpd.conf and restarted Apache:

[Megunticook]

Update: had another brief server downtime, but it came back up by itself after 3 minutes. I upgraded my AWS instance to 8GB memory and revised my httpd.conf settings to set MaxRequestWorkers and ServerLimit to 75. Everything seems to be fine now, 6.7GB memory free, 111 total tasks, 1 running, 69 sleeping.

I think the MaxRequestWorkers and ServerLimit were too low. Here's the last few log entries from httpd_error:

[James Blond]

That are very high settings. Unlike prefork event and worker mpm start only a few processes and those start threads. The threads then handle the requests.

Here are the stettings I work with.

[Megunticook]

Thanks for the clarification. I'll try lowering those--was just following the formula from that AWS blogpost. Yesterday I removed the directive on MaxWorkerRequests and MaxServers from my httpd.conf just so Apache would use the default values. I'll try using lower values next.

Today I saw another head-scratcher in my log file. Did some Googling but if you have any insights to what is happening here I'd really appreciate it. Here's the most recent entries in this morning's log:

[James Blond]

[Megunticook]

That's odd...I just checked my configuration saw heartbeat was already commented out by default: