very high load on XEN VPS creation

Hi,

I can't create XEN VPS instances on one of my servers. The load get so high that that the server crashes.

The server runs:

root@usaxen03:[~]$ cat /etc/redhat-release

CentOS release 5.4 (Final)

root@usaxen03:[~]$ uname -a
Linux usaxen03 2.6.18-164.el5xen #1 SMP Thu Sep 3 04:03:03 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux

Here's an output from top:

top - 13:45:24 up 13:34,  3 users,  load average: 68.26, 65.07, 57.32
Tasks: 206 total,   4 running, 199 sleeping,   1 stopped,   2 zombie
Cpu(s):  0.1%us,  7.7%sy,  0.0%ni,  0.0%id, 92.1%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:    262144k total,   258828k used,     3316k free,      232k buffers
Swap:  8191992k total,  7491540k used,   700452k free,     3388k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
  194 root      10  -5     0    0    0 D  0.9  0.0   0:56.34 kswapd0
16766 root      18   0 33472 1904  548 D  0.6  0.7   0:02.14 bandwidth.pl
16793 root      18   0 93804  472  316 R  0.6  0.2   0:00.62 crond
16730 root      18   0  109m 1052  452 D  0.6  0.4   0:03.93 miniserv.pl
16759 root      18   0 33472 1972  548 D  0.6  0.8   0:03.33 bandwidth.pl
16738 root      18   0 38156 1608  588 D  0.5  0.6   0:03.69 status.pl
4729 root      15  -1  112m  428  340 S  0.5  0.2   0:08.13 modclusterd
16765 root      18   0 34248 1492  584 D  0.5  0.6   0:02.31 status.pl
16798 root      18   0 74852  184  108 D  0.5  0.1   0:00.07 crond
2332 root      17   0 62612  428  376 D  0.4  0.2   0:00.23 sshd
16464 root      16   0 12740  956  644 R  0.4  0.4   0:05.93 top
16739 root      18   0 76704  260  120 D  0.4  0.1   0:03.83 status.pl
16740 root      18   0 76704  468  172 D  0.4  0.2   0:03.49 status.pl
16748 root      18   0 76704  268  120 D  0.4  0.1   0:02.95 status.pl
16784 root      18   0 60436  404  224 D  0.4  0.2   0:01.40 sshd
4893 ricci     15  -1 55936  344  332 S  0.3  0.1   0:01.26 ricci
16374 root      18   0  111m 1604  372 D  0.3  0.6   0:06.35 miniserv.pl
16615 root      18   0 83248 2104  760 D  0.3  0.8   0:05.55 bandwidth.pl
16745 root      18   0 76704  260  120 D  0.3  0.1   0:03.70 status.pl
16770 root      17   0 56164  816  580 D  0.3  0.3   0:01.93 ssh
16773 root      18   0 33472 2644  752 D  0.3  1.0   0:01.74 bandwidth.pl
16776 root      18   0 60520  728  484 D  0.3  0.3   0:02.00 sshd
16388 root      18   0 74188 1912  628 D  0.3  0.7   0:06.26 bandwidth.pl
16529 root      18   0 53644 1704  456 D  0.3  0.7   0:05.24 fastrpc.cgi
16530 root      18   0 75376 1928  864 D  0.3  0.7   0:05.10 bandwidth.pl
16645 root      18   0 76704  924  736 D  0.3  0.4   0:02.91 status.pl
16735 root      18   0  121m  608  276 D  0.3  0.2   0:03.39 lfd
16747 root      18   0 76704  272  120 D  0.3  0.1   0:03.73 status.pl
16748 root      18   0 76704  268  120 D  0.1  0.1   0:02.89 status.pl
16759 root      18   0 33472 1828  496 D  0.1  0.7   0:03.24 bandwidth.pl
16766 root      18   0 33472 1888  488 D  0.1  0.7   0:02.04 bandwidth.pl
16773 root      18   0 33472 2564  500 R  0.1  1.0   0:01.69 bandwidth.pl
16774 root      18   0  121m  400  116 D  0.1  0.2   0:02.42 lfd
16776 root      18   0 60520  764  520 D  0.1  0.3   0:01.95 sshd

It's clearly using way too much swap, but I can't understand why.

the server rhas a Dual Core 3Ghz CPU, 2GB RAM and SATAII HDD. Other servers with similar configuration didn't give me this problem.

Status: 
Closed (fixed)

Comments

How much RAM does your Xen host system have? From that top output, it looks like only 256M, which is pretty low .. certainly too low to host a Xen guest.

The server has 4GB RAM,

xentop - 21:02:03 Xen 3.1.2-164.el5 2 domains: 2 running, 0 blocked, 0 paused, 0 crashed, 0 dying, 0 shutdown Mem: 4118360k total, 634332k used, 3484028k free CPUs: 2 @ 2997MHz

and have 256MB limited to dom0 in grub.conf

title CentOS (2.6.18-164.el5xen) root (hd0,0) kernel /xen.gz-2.6.18-164.el5 dom0_mem=256M module /vmlinuz-2.6.18-164.el5xen ro root=/dev/fluid02/root module /initrd-2.6.18-164.el5xen.img

AFAIK, top doesn't report the full RAM on XEN systems.

Does CloudMin need the full RAM to be accesible to dom0 ?

Perhaps 256MB is too low, causing thrashing during the disk-intensive Xen instance creation process.

You might want to up that to 512MB if possible.

I have removed the limit, rebooted and successfully created a new VPS. Thanx for the help :)

Great!

I would be interested to see if this helps with either of the other issues you reported - often low memory can cause odd and hard-to-diagnose failures, as it causes commands to mysteriously fail.

are you referring to the LVM & swap issues? This problem is on usaxen03, and those errors are on usaxen05 - so it's not related :)

Ok .. but if usaxen05 also has only 256MB of RAM for domU, you might want to up that as well.

I'll keep this to the other topic, to make it a bit easier :)

Automatically closed -- issue fixed for 2 weeks with no activity.