Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Help Me Understand High Load on Openvz Container
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Help Me Understand High Load on Openvz Container

GM2015GM2015 Member
edited November 2015 in Help

I've had my $10/year(what I paid) or as $1/m as it's known server with dacentec suspended for high load(according to their email) for a few days. I've asked it to be reactivated, which they did quickly(less than 5 minutes).

All I've got is 2 wordpress sites on the container. Both of them using the same set of plugins.

Plugins:
w3tc, google sitemaps, disable google fonts, unnamed.

The unnamed plugin seems to be cpu and ram intensive sometimes from my experience with it on digitalocean, where using it isn't a problem.

This is NOT a complaint thread, so please take your hate and sense of urgent need to defend dacentec elsewhere. I've got no right to complain and I'm NOT doing that. Understand?

I just noticed today that the server was turned off(it got apparently turned off again after 14 hours). Earlier I've read dacentec's http://www.lowendtalk.com/discussion/comment/1384073/#Comment_1384073 reply to their http://www.lowendtalk.com/discussion/67802/dacentec-1-vps-are-now-re-stockednew thread. I realized that the server being turned off meant I again hit their "abuse"(I hate this word in this sense) definition.

Back to the question, I tried reactivating the plugins I use on the server(caching plugin plus something else) and I noticed that the container's load quickly increased to about 2 while the wordpress site became unresponsive(browser displaying waiting for domain.com).

I quickly killed the webserver and deactivated all plugins via mysql.

Apparently, on reactivating w3tc, cpu usage increases a few %. Wait time is also high, which from what I've read here is a sign of expected heavy overselling.

Top:

top - 18:37:57 up 35 min,  1 user,  load average: 0.94, 0.40, 0.35
Tasks:  25 total,   1 running,  24 sleeping,   0 stopped,   0 zombie
%Cpu(s):  1.0 us,  2.0 sy,  0.0 ni, 90.9 id,  6.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem:    524288 total,   519904 used,     4384 free,        0 buffers
KiB Swap:   262144 total,    18448 used,   243696 free.   488696 cached Mem

  PID USER      PR  NI    VIRT    RES    SHR S %CPU %MEM     TIME+ COMMAND      
  779 1001      20   0  103692  18624  11948 D  2.3  3.6   0:03.24 php5-fpm 

While I am in no way complaining and don't even take it this way, but why is the load at 0.94 when all I try is to reactivate the caching plugin?

I don't understand why the load becomes high so quickly. 4 sites are running on digitalocean's 512MB plan and there's barely any load(0.00-0.20). But I've separated mysql and the nginx+php5fpm on digitalocean.

Strange. I don't like openvz.

Comments

  • How many wordpress posts do you have?

    Thanked by 1GM2015
  • Quite a few.

    DalComp said: How many wordpress posts do you have?

  • JacobJacob Member
    edited November 2015

    I/O wait is not a high amount by any means (6.0), while this number should be relatively low any way, even on full nodes... I'd start getting worried when you're constantly in double figures.

    I'd check the usual, max_execution time, input_time along with your timeouts and give your memory limit a glance... What's the recommendations for the plugin?

    I've never used that plugin before, but then again I'm not a fan of Wordpress anyway.

    Thanked by 1GM2015
  • The caching plugin(w3tc) runs very well on digitalocean. I don't understand why reactivating it causes the site to be frozen for so long.

    Jacob said: What's the recommendations for the plugin?

  • I've separated mysql and the nginx+php5fpm on digitalocean.

    So on DO you had 2 servers, and now you have one?
    Have you checked what the SQL server load is like at DO?

    Thanked by 1GM2015
  • @GM2015 said:
    Quite a few.

    I managed a wordpress site with thousands of posts, w3tc basically cache each of these pages, and create multiple small files for each of these. These files are generated every 180 seconds (default settings), and thus if you have a lot of posts, I/O could get pretty busy (as seen in your increased %wa). DO is SSD thus they can handle it better compared to HDD.

    Try to increase these settings, i.e. "Maximum lifetime of cache objects", especially on object and page cache section of w3tc. You could also try to use memcache.

    Thanked by 1GM2015
  • GM2015GM2015 Member
    edited November 2015

    No, what I've tried to mean is, I've got four other sites running on 2 digitalocean droplets. 3 of those sites use about the same set of plugins and do the same tasks as the 2 site on the dacentec container.

    The difference is, on digitalocean, I've separated the webstack into a mariadb droplet and a nginx+php5-fpm droplet.

    The mysql load is very low, but the mysql server I think almost fully uses its allocated ram.

    bohdans said: So on DO you had 2 servers, and now you have one?

  • About the small files, that's damn true. I tried to remove all caches of previous activity(before suspension), so that I can restart using w3tc from scratch again with my settings saved. It took minutes to delete all those I don't know how many files.

    W3tc also used up all inodes on my digitalocean webserver node. I cleared all crap of w3tc and inode usage went down from 100% to about 5%.

    DalComp said: These files are generated every 180 seconds (default settings), and thus if you have a lot of posts, I/O could get pretty busy (as seen in your increased %wa).

  • @GM2015 said:
    About the small files, that's damn true. I tried to remove all caches of previous activity(before suspension), so that I can restart using w3tc from scratch again with my settings saved. It took minutes to delete all those I don't know how many files.

    W3tc also used up all inodes on my digitalocean webserver node. I cleared all crap of w3tc and inode usage went down from 100% to about 5%.

    Yeah, I remember the site (and a few similar sites) reached over 2.5 million inodes
    http://www.lowendtalk.com/discussion/12069/inode-allocation#latest

    That and the high I/O issue made me jumped several providers before I could find the real culprit.

    Thanked by 1GM2015
  • GM2015 said: digitalocean's 512MB plan and there's barely any load(0.00-0.20)

    Look at the difference in CPUs and storage.

    DO = E5-2630L + SSD storage

    There is more performance per core, so 100% utilisation on E5-2630L is a lot more activity than 100% utilisation on something like an L5420.

    Also with their high-speed SSD storage, IO is not so contended.

    Thanked by 1GM2015
  • IkoulaIkoula Member, Host Rep

    Hello,

    Just an idea :
    Maybe too much trafic ? could be attacks too ?
    Did you check number of vsitors/connections ?

  • Mahfuz_SS_EHLMahfuz_SS_EHL Host Rep, Veteran
    edited November 2015

    At DO, you're paying $5/Month, at Dacentec, it's less than $1/Month, how can the Performance be same ? And, there are few people including me whom I think haven't logged in the VPS bought from Dacentec because I don't manage any time to play with it. So, if you're lucky, you could have got that type of node which hosts people like Me :P Probably, you're not that lucky :) l

  • No, I didn't check for bots and crawlers. Analytics don't show too much javascript enabled visitors, so traffic's low on the container. Plus, it's been off for days because I didn't notice the suspension/turn off event until days. I've been wasting time for 2 weeks unfortunately.

    Ikoula said: Just an idea : Maybe too much trafic ? could be attacks too ? Did you check number of vsitors/connections ?

    We're talking about roughly similar specs:
    512MB Ram
    Different CPUs
    SSD vs HDD

    and while obviously DO is better in this scenario with a higher price, I didn't expect the container to have such loads and get suspended/turned off twice(1+1).

    Whatever. I turned off the plugin which was most likely responsible for the load and have reenabled w3tc caching from scratch, so loads almost nothing now.

    Hopefully the container won't be turned off if anything hits the 2 sites.

    Mahfuz_SS_EHL said: At DO, you're paying $5/Month, at Dacentec, it's less than $1/Month, how can the Performance be same ?

Sign In or Register to comment.