Re: Debian 7 + Asterisk + Virtualbox excruciatingly slow after upgrade
Hi
On Mon, May 13, 2013 at 07:25:34PM +0100, Ernie Dunbar wrote:
> We have a Debian virtual machine (hardware: Dual-core Xeon at 2.4Ghz, 4
> GB RAM, 2 GB of which is dedicated to the Debian VM) that's been running
> our Asterisk PBX for some time. Before Wheezy came out, it was doing
> reasonably well at its job, but now that we've upgraded to Wheezy, it's
> ridiculously slow, with load averages topping 15 on a regular basis.
> Naturally, at this server load, doing anything tests one's patience.
>
> Is this due to the new kernel, or the new version of Asterisk? The
> Asterisk version went from 1.6 to 1.8, and I recall that 1.8 is more of
> a resource hog than the previous version... however, we also have
> another Debian server (running Squeeze) with Xeon processors (4 cores in
> total, but the same generation of Xeons), and Asterisk 1.8 never uses
> more than about 40% of *one* CPU.
>
> I have few other clues as to what might be causing this problem,
> although the PHP upgrade also seems to be causing other, unrelated
> issues. Can anyone help me troubleshoot the problem better than I
> already have?
As others have mentioned, top is a good one to start with: keep an eye
on the CPU percentages.
- if %us +%sy is near 100%, point towards a cpu-bound system. Apart
from adding more cpu, optimising stuff, making the hamster wheel
spin faster or reducing load, there isn't much you can do. If the
system/application handled the same load before and now doesn't,
something really has become unoptimised.
- If free memory is low[1], buffers are low[1], load average is high
and %wa is high >15% [2], this points towards a memory constrained
system. Use vmstat or similar to see whether you're swapping - this
would be indicated by nonzero activity in the "si" and "so"
columns. If so, reduce memory usage or add memory.
- If free memory is OK, %wa (wait-for-io) is > 15% [1] then you most
likely have an I/O bound system, i.e. with apps performing IO rather
than swapping. Use vmstat to conform - this would be indicated by
zero activity in "si"/"so" columns, but large numbers in "bi"/"bo".
That should cover the basics as it appears to apply to your
system. Tools like "top" and "vmstat" are your friends, although it is
far from obvious how to interpret the output.
Hope this helps
[1] "low" is a deliberately fuzzy term here. E.g. less than 100Mb
free on a 2G system sounds reasonable.
[2] The exact threshold for when %wait-for-io is "high" is highly
debatable and varies between use cases.
>
>
> --
> To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
> Archive: [🔎] 04a6c2cf17ff4f8aad8ecac4de715c58@pop.lightspeed.ca">http://lists.debian.org/[🔎] 04a6c2cf17ff4f8aad8ecac4de715c58@pop.lightspeed.ca
>
--
Karl E. Jorgensen
Reply to: