Need assistance troubleshooting PS memory related instability


#1

Hi,

I have a PS that we have been using for several years. For the longest time we have had it running memcron and consistently using minimum memory…

Over the past couple of months, the memory usage has been causing the PS to give us regular warnings about instability, and the PS is resetting itself.

If you are familiar with this, you will know the PS usage reports in the panel do not help since they do not record what happens at the time of the ‘overload’ and overall usages seems consistently below the minimum PS setting. Even disabling memcron and doubling this to 600mb leaves me with stability emails and PS reboots.

I have looked at TOP for a long while and never see amazing spikes, however I do see 5-10% memory usage per domain on the PS - still not enough to go over the limit - and certainly not since I understand the PS also has a burst limit for short intervals…

I have followed some of the instructions in the Wiki to try to ascertain what is the cause - but I do not see anything Obvious - and do not have more time to dig (time to learn where to dig, how to dig and then do the digging!)

At this point I am about to move away from DH because I cannot understand what I need to do other than jack up the memory (and the price) to levels that do not seem necessary for our usage.

Before I do this, I would like to see if anyone on the board who is familiar with DH PS system, and maybe has some experience solving this themselves, could assist. It could be as simple as adjusting a couple of settings, or it could be something more involved like stitching webservers (I have read the apache is not very parsimonious in its memory use, and something like lighttpd is a better choice?)

Since it will cost me a lot of time to move, or more money in PS fees to keep pushing the memory up, I can justify paying for this service, but am really not sure what it will take (hopefully not too much if its is done by someone who knows their stuff and can zero in on the problem fast!).

If you think you can help - and have an idea how you would charge for this, please PM me your contact details and I will be in touch to work out how to proceed.

Many thanks!

Alex
(PS: I am located in Singapore - and if necessary can talk via skype - but do bear in mind the time difference since we are on the other side of the world from any experts in the US. Late night communication is not a problem on our side though)


#2

Try asking support to turn off your nightly access logs from being generated. It could be the culprit.


#3

Thanks for that Perry! I’ll try asking them for that now! Lets see what happens… I’ll post any changes.

In the meantime…

Even if Perry’s suggestion fixes things, I still would like to have another set of experienced eyes on our PS for a little while - since this is a relatively new issue and may not be entirely DH’s scheduled jobs problem :wink:

I have 20+ domains registered on the PS, but many are mirrors and/or dormant. I have 6 joomla installations (4 of which have no more than 3-4 users / day traffic, the other two, perhaps 100 visitors total in a day) and 1 drupal installation (still being installed/tested.) Also have a few other FOSS php progs (1x jobber board and 1x prestashop) that are rarely accessed (1-2 visits a day).

With PS memory set to 300mb TOP reports the busier sites as having 10-15% memory consumption and about 4-5 other domains with 5-6% each.
Total~ 50-60%
This (150-165mb) might be the ‘resting’ memory usage of the sites (ex: application-level caches), but I would expect that to scale sensibly per user, not to die on the very low volume i have described.

It would also be useful to work out who/what is the culprit (if its not Perry’s log generation) and see what can be done to optimize without trying to rewrite FOSS applications!

Still very curious…
Alex


#4

OK, The log generation turned out to be a dead end, (worth a try!) but it did get a response from DH support after I asked them to turn off the log generation. They recommended JJs memory manager for VPS. I have left this running and its doing a way better job that memcron that I had previously installed.
Problem is memory is bouncing all over the place and I would love to try to get a handle on who/what is causing it. Looking at the odd top is not resulting in anything meaningful for me, and JJs tool does have some process logs but no docs on how to interpret them and no data on which users (ie: domains) are using what memory.

If there a way to monitor this more effectively without ‘catching it in the act’? I would love to see a log of ‘average’ memory consumption over time - by domain or even by script/php page.

Can anyone think how I can find this out? At least it would give me a target to lock onto when working out what is wrong.
At present I just know something is off, and that its costing me $$$ to keep running with more than double of the memory we ever used before…

I’m still really open to an experienced person doing the digging in my place! Any takers? let me know how you would charge…


#5

Hi,

I am having the same issue. I am getting emails from DH saying that I am over the 300MB limit. But I am not running heavy at all, even less than your description. The graph shows maybe 10% of usage of the allotted 300MB, yet I am getting rebooted every week and notified by DH that I am exceeding the allocation. If you find the solution please post on the forum. Thanks!