SUMMARY: More Pine probs (SOL2.6/NFS)

From: Admin Staff (admin1@berkshire.net)
Date: Tue Jun 02 1998 - 16:26:53 CDT


Thank you to all who replied, I got several very insightful answers, which
all could be a part of the problem. I am leaning toward the following
reasons, provided to me by respondants: Checkpointing (pine checking for
new mail and writing changes in message status frequently), NFS V3 (it was
pointed out that pine has a difficult time recognizing that it's on NFS in
Solaris, and tries to lock the message between each write), and possible
dropped packets (although I think it's unlikely with new sun machines, new
sun enet cables and new cisco switches). Thanks to all who replied. Is
there a Debugging function to NFS that allows you to print summaries of
dropped or timed out packets?

I'm enclosing key responses and my original post below.

-Jason

On Tue, 2 Jun 1998, Ronald Loftin wrote:
>
> Date: Tue, 2 Jun 1998 08:36:45 -0400 (EDT)
> From: Ronald Loftin <reloftin@mailbox.syr.edu>
> To: Admin Staff <admin1@berkshire.net>
> Subject: Re: More Pine probs (SOL2.6/NFS)
>
>
> We have this problem here. The problem is that under NFS version 3 the
> file locking mechanism changes, and PINE doesn't know that it's using NFS.
> Pine tries to lock the mail file between each message write, which takes a
> noticeable amount of time via NFS. The interim workaround ( until Pine
> 4.0 is available ) is to use NFS Version 2 for filesystems where Pine mail
> files live. We too are waiting for the final fix, but it's going to have
> to come from the authors of Pine, since Sun has declined to do anything
> about it.
>
>
> Ronald Loftin E-mail: reloftin@mailbox.syr.edu
> Computing and Media Services Syracuse University
> Voice: 315-443-4103 Fax: 315-443-3817
>
>
>

On Mon, 1 Jun 1998, Michael Baumann wrote:
>
> Date: Mon, 1 Jun 1998 15:38:53 -0700
> From: Michael Baumann <baumann@proton.llumc.edu>
> To: Admin Staff <admin1@berkshire.net>
> Subject: RE: More Pine probs (SOL2.6/NFS)
>
> [The following text is in the "iso-8859-1" character set]
> [Your display is set for the "US-ASCII" character set]
> [Some characters may be displayed incorrectly]
>
> Yep, I can provide insight. It is checkpointing.. I have the
> same problem under 4.x as well.
>
> The fix is to keep your inbox < 500 messages.
> Move them suckers off to a seperate folder :)
>
>
> --
> Michael Baumann Optivus Technology Inc.|Loma Linda University Medical
> Center
> San Bernardino, California. (909)799-8308 |Internet: baumann@llumc.edu
>
>
>

On 2 Jun 1998, Heidi Burgiel wrote:
>
> Date: 2 Jun 1998 17:50:20 -0000
> From: Heidi Burgiel <burgiel@math.uic.edu>
> To: admin1@berkshire.net
> Subject: Re: More Pine probs (SOL2.6/NFS)
>
> Have you checked for dropped packets? We traced similar behavior in
> pine to timeouts waiting for dropped packets. Try shortening the
> timeout on your NFS transactions, if you haven't done so already.
>
> My apologies if you've already ruled this out.
>
> Heidi B.
>

And my original message....

---------- Forwarded message ----------
Date: Mon, 1 Jun 1998 17:42:41 -0400 (EDT)
From: Admin Staff <admin1@berkshire.net>
To: Sun Managers <sun-managers@ra.mcs.anl.gov>
Subject: More Pine probs (SOL2.6/NFS)
Followup-To: Admin Staff <admin1@berkshire.net>

Hello All,

A while ago, I had a problem with slow pine performance. As indicated in
my subsequent summary, someone kindly pointed out that there was a time
difference between the mail NFS server and the client running Pine.
Syncing the time helped a lot but there is still a very present and
annoying hang in Pine about every minute or two. To ensure clock
syncronization, I'm syncing my clocks on my NFS clients and server to the
NTP server of a Cisco Router.

For those of you who are new to the issue:

I am having severly annoying _long_ hangs in Pine which happen about once
every few minutes. The hangs come at random points, and usually consist of
a complete hang of Pine, with or without the "**PINE" indicator in the
left hand corner. Hangs come often while just opening a message (cannot
scroll), in the Message index, and during message composition. Are long
holdups are during message expunging and the closing of the mailbox. Hangs
last between 1 and 2 minutes.

I am running Pine on an NFS mounted mailbox, on an _under_ used Ultra1
with 256 megs of ram, mounting its mailbox over a switched 100Mbps full
duplex connection to a dual CPU ultra-2 with 512 Megs of ram. Both
machines are using Solaris 2.6 with the recommended patches. I'm using NFS
version 3.

I've run "truss" on the pine process while it hangs. It appears to be
writing the mailbox, message by message. It does this every time it hangs,
like some kind of regular cleanup. It takes a while, and performs this way
too often. I would think it would write out the mailbox once in a great
while, maybe even only at close. Despite the fact that I have 1,000
messages, I would also think that this would happen quickly, given the
fact that I'm using 100mbps fdup enet and wide scsi drives.

Can anyone provide any insight into this? I have provided a sample of the
output of "truss" below. TIA. As usual, I will post a summary.

13954: uname(0xEFFFDA00) = 1
13954: open("/var/mail/zone.lock.896736423.13954.titan",
O_WRONLY|O_CREAT|O_EXCL, 0666) = 7
13954: chmod("/var/mail/zone.lock.896736423.13954.titan", 0666) = 0
13954: close(7) = 0
13954: link("/var/mail/zone.lock.896736423.13954.titan",
"/var/mail/zone.lock") = 0
13954: stat("/var/mail/zone.lock.896736423.13954.titan", 0xEFFFDF68) = 0
13954: unlink("/var/mail/zone.lock.896736423.13954.titan") = 0
13954: open("/var/mail/zone", O_RDWR) = 7
13954: getpid() = 13954 [13940]
13954: fstat(7, 0xEFFFDE70) = 0
13954: utssys(0xEFFFDE58, 0x2C80003, USTAT) = 0
13954: fcntl(7, F_SETLKW, 0xEFFFDE30) = 0
13954: fstat(7, 0xEFFFEC98) = 0
13954: sigaction(SIGHUP, 0xEFFFE6E0, 0xEFFFE780) = 0
13954: sigprocmask(SIG_UNBLOCK, 0xEFFFE7B0, 0xEFFFE7A0) = 0
13954: sigaction(SIGUSR2, 0xEFFFE6E0, 0xEFFFE780) = 0
13954: sigprocmask(SIG_UNBLOCK, 0xEFFFE7B0, 0xEFFFE7A0) = 0
13954: sigaction(SIGINT, 0xEFFFE6E0, 0xEFFFE780) = 0
13954: sigprocmask(SIG_UNBLOCK, 0xEFFFE7B0, 0xEFFFE7A0) = 0
13954: sigaction(SIGTERM, 0xEFFFE6E0, 0xEFFFE780) = 0
13954: sigprocmask(SIG_UNBLOCK, 0xEFFFE7B0, 0xEFFFE7A0) = 0
13954: lseek(7, 4446893, SEEK_SET) = 4446893
13954: write(7, "\0\0\0\0\0\0\0\0\0\0\0\0".., 5656) = 5656
13954: lseek(7, 0, SEEK_SET) = 0
13954: writev(7, 0xEFFFEBA8, 4) = 889
13954: writev(7, 0xEFFFEBA8, 4) = 3444
13954: writev(7, 0xEFFFEBA8, 4) = 517
13954: writev(7, 0xEFFFEBA8, 4) = 2411
13954: writev(7, 0xEFFFEBA8, 4) = 1703
13954: writev(7, 0xEFFFEBA8, 4) = 1041
13954: writev(7, 0xEFFFEBA8, 4) = 2405
13954: writev(7, 0xEFFFEBA8, 4) = 1746
13954: writev(7, 0xEFFFEBA8, 4) = 1736
13954: writev(7, 0xEFFFEBA8, 4) = 1145
(once for each message)

-Jason



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:41 CDT