Page 1 of 2 12 LastLast
Results 1 to 25 of 30
  1. #1
    Join Date
    Dec 2010
    Posts
    39

    %iowait VPS question

    In the last few weeks I have been having a lot issues with timeouts on a Joomla website (120seconds, which is the default for IPCConnectTimeout / fcgid). The website on that server has roughly 200 visitors per day.

    I checked the load with sysstat / sar and noticed that for quite some a few hours per day the %iowait hovers around 60 to 80 (average per day is roughly 8).

    My question: Is there anything I can tune on my VPS (Centos 5.6) to deal with this situation? I don't think that increasing the timeout to 10 minutes would help, because users wouldn't wait that long anyway.

    Thanks for any ideas in advance!

  2. #2
    Join Date
    May 2006
    Location
    San Francisco
    Posts
    7,325
    200 visitors per day total? That isn't much at all. Are you out of memory?

    It's possible that your VPS node is being overloaded by other customers. May be worth contacting your host about the iowait.

  3. #3
    Join Date
    Dec 2010
    Posts
    39
    I have 2GB (4GB burstable) memory (typlically memory usage is 1GB or less). I contacted the host (of course) already quite a while ago, but believe this situation could last for a "longer period. So I am checking what I could do on my end to deal with this situation.

  4. #4
    Join Date
    May 2006
    Location
    San Francisco
    Posts
    7,325
    What processes are showing up with the following command during high iowait?

    watch -n 1 "(ps aux | awk '\$8 ~ /D/ { print \$0 }')"

  5. #5
    Join Date
    Apr 2011
    Posts
    74
    Are you running any exotic plugins for your joomla site that could create issues? If not I would say that you don't get what you pay for. Make this clear to your host. If things don't change soon, give to another host your money. 200 visitors/day is nothing, so you should have no issues at all even with a non optimized vps.

  6. #6
    Join Date
    Dec 2010
    Posts
    39
    Thanks a lot, I will try that.

    Quote Originally Posted by Orien View Post
    What processes are showing up with the following command during high iowait?

    watch -n 1 "(ps aux | awk '\$8 ~ /D/ { print \$0 }')"

  7. #7
    Join Date
    Dec 2010
    Posts
    39
    I just run the standard plugins and the database is very small.


    Quote Originally Posted by MBGear View Post
    Are you running any exotic plugins for your joomla site that could create issues? If not I would say that you don't get what you pay for. Make this clear to your host. If things don't change soon, give to another host your money. 200 visitors/day is nothing, so you should have no issues at all even with a non optimized vps.

  8. #8
    iowait is usually storage congestion

  9. #9
    Join Date
    Dec 2010
    Posts
    39
    Still waiting to try:

    Code:
    watch -n 1 "(ps aux | awk '\$8 ~ /D/ { print \$0 }')"
    But the iowait has been ok so far (today). By the way thanks for your responses!

  10. #10
    Join Date
    Jun 2003
    Location
    Spain
    Posts
    4,251
    One Joomla site? I'm running 90 of them on my 1GB KnownHost VPS and it never misses a beat.

  11. #11
    Join Date
    Dec 2010
    Posts
    39
    Yes, I know, my VPS is very inexpensive compared to knownhost. I guess you get what you pay for - sometimes :-)

  12. #12
    Join Date
    Dec 2010
    Posts
    39
    Ok, today the iowait was horrible again, but I was able to try the command.

    Code:
    watch -n 1 "(ps aux | awk '\$8 ~ /D/ { print \$0 }')"
    There doesn't seemt to be "one" program in that list. Every program seems to show up for a while and then disappear for a while.

    The %iowait is around 70. The problem typically starts at every full hour and stops after 10-20 minutes (or even longer on very bad days). It looks like some scheduled jobs cause this (none of my jobs apparently).

  13. #13
    Join Date
    May 2006
    Location
    San Francisco
    Posts
    7,325
    Quote Originally Posted by 123customer View Post
    Ok, today the iowait was horrible again, but I was able to try the command.

    Code:
    watch -n 1 "(ps aux | awk '\$8 ~ /D/ { print \$0 }')"
    There doesn't seemt to be "one" program in that list. Every program seems to show up for a while and then disappear for a while.

    The %iowait is around 70. The problem typically starts at every full hour and stops after 10-20 minutes (or even longer on very bad days). It looks like some scheduled jobs cause this (none of my jobs apparently).
    This would suggest that there's heavy IO activity going on your node. Unfortunately, this is out of your control so it's something your host needs to investigate.

  14. #14
    Join Date
    Dec 2010
    Posts
    39
    Ok, understand.

    The host has been "investigating" for a few weeks already ... So hopefully they are able fix the problem (once they actually moved my OpenVZ container to a different node, but it didn't help).

  15. #15
    Quote Originally Posted by 123customer View Post
    Ok, understand.

    The host has been "investigating" for a few weeks already ... So hopefully they are able fix the problem (once they actually moved my OpenVZ container to a different node, but it didn't help).
    It sounds like they're just piling on a ton of customers onto nodes that aren't powerful enough to handle the customers on them. A couple hundred visitors a day isn't enough for your usage to be causing any problems. The only other potential cause is that a couple abusive users on the node are using up all the disk i/o, but if that were the case, the host should resolve that by suspending the abusive users. The fact that the problem persists certainly is a case of you get what you pay for: if the host were on the ball, they'd either have servers that were powerful enough to handle their customers, or they would investigate and suspend abusive customers (depending on the situation).
    IOFLOOD.com -- We Love Servers
    Phoenix, AZ Dedicated Servers in under an hour
    ★ Ryzen 9: 7950x3D ★ Dual E5-2680v4 Xeon ★
    Contact Us: sales@ioflood.com

  16. #16
    Join Date
    Mar 2011
    Posts
    659
    I bet you could throw the site on a ramdisk if you wanted a quick fix. If they're overselling they may not be thrilled by it, but you do what you gotta do
    Ain't here to spam my signature, I'm not desperate for sales.

  17. #17
    Join Date
    May 2009
    Location
    London, United Kingdom
    Posts
    472
    Who is your VPS provider? Do you know what drives are they using? Any RAID?

    You can test the I/O using the following command:
    dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

    Please share the output with us so we can tell you if it's acceptable.
    KnownSRV.com - Privacy. Managed. Secure. Guaranteed!
    Web Hosting | Dedicated Cloud VPS | Dedicated Server
    YOUR Day and Night, Fully Managed Hosting Solutions with REAL 24/7 Support

  18. #18
    Join Date
    Dec 2010
    Posts
    39
    Thanks, I will research this. Haven't done theat yet, any recommendations for Centos?

    Quote Originally Posted by jarland View Post
    I bet you could throw the site on a ramdisk if you wanted a quick fix. If they're overselling they may not be thrilled by it, but you do what you gotta do

  19. #19
    Join Date
    Dec 2010
    Posts
    39
    I don't want to name the provider in this thread. I am just trying to figure out some options how to deal with the situation. I never expected high performance and 99% availability from them. Just a good compromise between price and performance. It had worked until april ...

    I believe they mentioned Raid 10 once.

    And here is the result for your test:

    Code:
    [root ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 111.718 seconds, 9.6 MB/s
    Quote Originally Posted by SceneSRV View Post
    Who is your VPS provider? Do you know what drives are they using? Any RAID?

    You can test the I/O using the following command:
    dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

    Please share the output with us so we can tell you if it's acceptable.

  20. #20
    Join Date
    May 2009
    Location
    London, United Kingdom
    Posts
    472
    Ouch, it's TERRIBLE. I'd ask them to migrate you to some other VPS Node, if they have any Node that is NOT overloaded like your current one(from the looks off your output).
    KnownSRV.com - Privacy. Managed. Secure. Guaranteed!
    Web Hosting | Dedicated Cloud VPS | Dedicated Server
    YOUR Day and Night, Fully Managed Hosting Solutions with REAL 24/7 Support

  21. #21
    Join Date
    Dec 2010
    Posts
    39
    Thanks for the feedback. What value would be acceptable?

    I asked them to move my container again, but they rejected this. They said this would have a significant negative impact on the iowait. Not sure whether I should push for this (because the other nodes had similiar issues before).

    I should probably try the RAM disk, my site is very small (20MB).

  22. #22
    I don't want to name the provider in this thread
    Most people will assume it based of your user name, regardless if its correct or not.

  23. #23
    Quote Originally Posted by 123customer View Post
    What value would be acceptable?

    To compare, this is a single 7200rpm SATA hard drive:

    Code:
    [root@nlweb ~]# cd /hdd2
    [root@nlweb hdd2]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 9.63441 seconds, 111 MB/s

  24. #24
    Join Date
    Jan 2003
    Location
    Canada
    Posts
    5,073
    Quote Originally Posted by 123customer View Post
    Thanks for the feedback. What value would be acceptable?

    I asked them to move my container again, but they rejected this. They said this would have a significant negative impact on the iowait. Not sure whether I should push for this (because the other nodes had similiar issues before).

    I should probably try the RAM disk, my site is very small (20MB).
    I should add that ramdisks are a scary thing if your node reboots you'll lose the data that was in there requiring you to refill it.

    You could probably have any sort of caching folders go there to help some but even then most decent webservers will cache to memory some at least anyways.

    Francisco

  25. #25
    Join Date
    May 2011
    Location
    India
    Posts
    106
    it looks the node is oversold. contact you host to move to another node or switch the host.

Page 1 of 2 12 LastLast

Similar Threads

  1. iowait question
    By imho in forum Hosting Security and Technology
    Replies: 6
    Last Post: 06-30-2010, 11:47 PM
  2. iowait
    By elas in forum Hosting Security and Technology
    Replies: 42
    Last Post: 09-30-2009, 03:37 AM
  3. high bandwidth advice needed. and an iowait question...
    By disgust in forum Dedicated Server
    Replies: 28
    Last Post: 11-07-2005, 09:18 PM
  4. Anyone with 2.6? iowait test
    By MattF in forum Hosting Security and Technology
    Replies: 4
    Last Post: 08-16-2004, 12:48 AM
  5. Serious iowait issues
    By ezclassified in forum Hosting Security and Technology
    Replies: 1
    Last Post: 06-26-2004, 02:43 PM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •