Never ending i-framp in polymix-4

From: Joe Cooper (joe@swelltech.com)
Date: Tue Nov 09 2004 - 16:14:51 MST


I've built a new Polygraph environment this week, running Polygraph
version 2.8.1 on two Fedora Core 1 Linux boxes. Everything built and
installed without trouble, and everything runs without errors, including
generation of IPs and generation of the DNS zone files. (Hooray! I
remembered it being harder a couple of years ago when I setup my last
polygraph pair.)

I was able to run simple.pg locally and across tes boxes and a
datacomm-1.pg with the proxy in the loop. All of them worked without
trouble.

However, when I run polymix-4.pg, it seems to get stuck in the "i-framp"
fill phase. I've specified "size ProxyCacheSize = 10GB;", and so I
expected it to stop at 20GB and move into the next phase, but the last
status printout five hours into the run shows fill.size at 34.70GB and
still growing:

295.21| resource usage:
         CPU Usage: 3.28hour sys + 30.95min user = 3.80hour
         Page faults with physical i/o: 692

295.21| fyi: phase progress:
         duration: 4.92hour goal: 20.00min (1475.03% complete)
         xact.count: 4712964
         fill.size: 34.70GB
         xact.errs.count: 0
         xact.errs.ratio: 0.00
         WSS frozen

295.21| fyi: min 'direct' objects in working set:
         global public: 510641 (~5.17GB size, 569/609=93.43% frozen slices)
         local private: 169626 (~1.72GB size, 391472/391472=100.00%
frozen slices)

WSS froze about three hours ago at 5.27GB, and fill.size crossed 20 GB
soon after. I'm using the polymix-4.pg that shipped with this version
of polygraph, so it has the updates for robot interests discussed here,
so I guess is isn't a known phase sync issue:
http://www.web-polygraph.org/downloads/upgrade2p8p0.html

The only changes I've made to the workload are the IPs of my test
machines, the fill rate, the resolvers, and the fill rate.

What else am I missing?

Thanks!



This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:28 MST