reader performance issue
tony.mills at millportsystems.co.uk
Wed Mar 16 19:47:03 UTC 2005
I am having a problem and wondered if you would be able to point me in the right direction.
I am doing a migration project of a news server farm. The old system is running inn 2.3.2, it runs on solaris 8 servers. we have one master server and 3 reader servers that fire up nnrpd under xinetd. The backend is being provided from netapp filers (NFS).
the new system has exactly the same setup but we are using inn 2.4.1., the master server and the netapp filers are more powerful, and we have more storage.
when migrating we also increased the size of our (CNFS) cyclic buffers from 2gb to 50gb as it was believed that this would improve nnrpd startup time. after going live the new platfform performed very badly, for a client to serv 8Mb of traffic it flatlined our 100Mb interface on the backend with data coming from the nfs filer..
we do not see this symptom on the old system.
we have nfsreader and nfswriter turned on but even without these settings it doesnt make any difference.
rsize and wsize are both 32768 at both ends (nfs server and unix client)
I am hoping someone else has seen this and may have a way of resolving the issue.
I have read a reference about large cnfs filesystems becoming slower as they get bigger, is that the case here? it would be a pain but it would be possible to go back to >1000 2gb files if we needed to. It would be great to get some input from the the developers before it do it.
when i emulate <>180 users downloading random articles the server spins with dnlc_purge's but i think thats because the backend interface is overloaded more than an ncsize issue.
thanks in advance
More information about the inn-workers