[9fans] fossil server crashing

[9fans] fossil server crashing

[Andrew]

As the subject says, my fossil server keeps locking up. It is most easy to
duplicate by running the cpu command from a terminal. The disk activity
light on the main hard drive lights up, and stays lit. If i run cpu from
a drawterm connected to the auth server, to connect to the fossil server,
i get the same results but the cryptic message "msgWrite: delay r Rflush
tag 7" when i hit delete on the window. The 7 is actually a number from
somewhere around 3 to usually 40

Everything, including the fossil server mounts their root off the fossil
drive, which happens to be a scsi on a sd53xxx controller.

Any ideas?

this is mostly a recent thing since i pulled from sources

Re: [9fans] fossil server crashing

[Russ Cox]

your subject says fossil is crashing.
your text says fossil is hanging.

the message about Rflush is harmless.
it's just telling you that the flush
implementation actually works.

can you hear disk activity?
perhaps there's a problem with
the disk driver.

fossil hasn't changed in months.

Re: [9fans] fossil server crashing

[Andrew]

On Mon, Jun 02, 2003 at 11:41:13PM -0400, Russ Cox wrote:
> your subject says fossil is crashing.
> your text says fossil is hanging.
sorry, i was in a hurry, i have a broader definition of crashing...anyways
>
> the message about Rflush is harmless.
> it's just telling you that the flush
> implementation actually works.
i thought anything would be helpful
>
> can you hear disk activity?
> perhaps there's a problem with
> the disk driver.
the light comes on, thats it, i cant _hear_ any activity

im trying to swap over to a nice simple ide drive, but its hard when I
cant keep the fossil server up for more than a few minutes
>
> fossil hasn't changed in months.
>

Re: [9fans] fossil server crashing

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 506 bytes --]

Do you have your kernel sources uptodate wrt sources?
Long ago there was a problem with the scheduler that allowed fossil
to eat the cpu (because of a sleep(0) IIRC).

I also run a fossil updated from sources and it runs just fine.
Have you double checked diff'ing w/ sources? (I'd check fossil and kernel).

Can you try using a fossil that runs off a different disk? (Or perhaps
using a kfs on the disk and starting a fossil running from a file), just
to see if it's a sd thing (I'm using ide).

[-- Attachment #2: Type: message/rfc822, Size: 1949 bytes --]

From: Andrew <afrayedknot@thefrayedknot.armory.com>
To: 9fans@cse.psu.edu
Subject: [9fans] fossil server crashing
Date: Mon, 2 Jun 2003 20:29:23 -0700
Message-ID: <20030603032923.GA2474@thefrayedknot.armory.com>

As the subject says, my fossil server keeps locking up. It is most easy to
duplicate by running the cpu command from a terminal. The disk activity
light on the main hard drive lights up, and stays lit. If i run cpu from
a drawterm connected to the auth server, to connect to the fossil server,
i get the same results but the cryptic message "msgWrite: delay r Rflush
tag 7" when i hit delete on the window. The 7 is actually a number from
somewhere around 3 to usually 40

Everything, including the fossil server mounts their root off the fossil
drive, which happens to be a scsi on a sd53xxx controller.

Any ideas?

this is mostly a recent thing since i pulled from sources