[9fans] venti woes

[9fans] venti woes

[Steve Simon]

Hi,

I have been foolish and broken my machine :-(

This is a standalone terminal which used to have
a ide kfs and I am migrating to SCSI and fossil+venti.

I installed the scsi disk from the bootable CD and copied
my user area accross trivially.

Then I formated the venti arenas and index, initialised the index,
and set venti=#S/sd00/venti in plan9.ini - and rebooted

All went fine for a while but after an hour or so the disk started
rattling and the machine hung (Rio completely hung, ^T^T^P still worked).

I was very surprised as the snaptime specified the snap -a to happen at 3am
(see below).

I rebooted and the disk rattled again but this time rio was responsive
so I ignored the disk and worked on. eventually the disk stopped.

Somtime later I noticed my mailbox has become a zero length file.

I rebooted again and now my profile has become a zero length file.

My theroy is that fossil is sending the data to venti but seems unable
to fetch archived files from it.

I kept the IDE disk "just in case" so its not so aufull, and I could
reinstall, but I I would probably do the same things and get the
same result.

	anyone any ideas?

-Steve


larch% venti/conf /dev/sd00/fossil
index main
isect /dev/sd00/index
arenas /dev/sd00/arenas

larch% fossil/conf /dev/sd00/fossil
fsys main config /dev/sd00/fossil
fsys main open -c 3000
fsys main snaptime -a 0300 -s 15 -t 1440
listen tcp!*!564

Re: [9fans] venti woes

[andrey mirtchovski]

i've had this happen with large snapshots to venti -- the system will
hang completely (can't open a new window in rio) but apps already
started, such as stats, will continue working.  it goes away after the
'snap -a' and the system becomes usable again...

maybe you need to leave the machine be for the first big snapshot.
why does it happen out of sync (not at 3am) is an interesting question
though :)

andrey