Client side stats SSL workload 4.3.2

Alex Rousskov rousskov at measurement-factory.com
Thu Jun 6 22:50:32 UTC 2013


On 06/05/2013 01:31 AM, Michael Hendrie wrote:

> Finally back to looking at this and some analysis seems to indicate a slow memory leak for SSL tests when sent to a forward proxy in 4.3.2.  In direct HTTP(S) tests (no forward proxy) and  HTTP -> Forward proxy test the memory usage off polygraph-client remains fairly stable once the tests settles into it's top phase.  In HTTPS -> Forward proxy however memory usage of polygraph-client continues to grow and eventually test fails

> Any ideas of where to look given the above analysis?  My C++ skills are fairly weak so not sure what assistance I can be in tracking but happy to look.

You can try running Polygraph under valgrind control and analyzing
valgrind reports, but that requires some C++ skills and Polygraph
knowledge to weed out false positives. I suggest filing a bug report for
this problem: https://bugs.launchpad.net/polygraph

If you can, please include the workload you are using and some kind of
log showing memory usage growth.

We will try to reproduce and, if needed, fix the leak.


Thank you,

Alex.




More information about the Users mailing list