Increasing transient memory usage

Bryan Stillwell bstillwell at photobucket.com
Mon Aug 22 04:58:21 CEST 2011


I believe I'm running into the problem mentioned in this bug report:

https://www.varnish-cache.org/trac/ticket/953


I've tried the following fix, but it doesn't appear to fix the problem:

https://www.varnish-cache.org/trac/changeset/beb0c5b1f4f49d711822e90ca73d69bbed683a71


The behavior I'm seeing is that once SMA.s0.nbytes hits the number
specified in '-s malloc,?G' the value of SMA.Transient.nbytes starts
going up (it's at zero the whole time before hitting the '-s
malloc,?G' number).

I have two machines I'm testing with that both have 32G of memory
(varnish 3.0.0 on centos 6.0), but for troubleshooting this problem I
set one to use 4G and the other to 8G.  The values after running for
around an hour are:

4G test config:
SMA.s0.nbytes         4294213238          .   Outstanding bytes
SMA.Transient.nbytes  3137682125          .   Outstanding bytes

8G test config:
SMA.s0.nbytes         8589339844          .   Outstanding bytes
SMA.Transient.nbytes  1260306015          .   Outstanding bytes


So the 4G config is using less memory after an hour, but still quite a
bit more than the 20% that is recommended to leave for overhead (plus
it continues to climb).

Is anyone besides the people on that bug seeing this problem?  Are
there any other patches or workarounds?

Thanks,
Bryan




More information about the varnish-misc mailing list