Re: [Somewhat OT] Curiouser and curiouser
- From: ibuprofin@xxxxxxxxxxxxxxxxxxxxxx (Moe Trin)
- Date: Sun, 06 May 2007 19:48:39 -0500
On Sun, 06 May 2007, in the Usenet newsgroup alt.os.linux.suse, in article
<Fvp%h.22$Qv7.17@xxxxxxxxxxxx>, Rajko M. wrote:
Moe Trin wrote:
I was wondering why on Ghodz green
earth SuSE even adopted the thing.
It was promissing alternative to other file systems.
I think I have to go along with Ted T'so here. The tradeoffs were not
worth it.
It has a HORRIBLE reputation for reliability.
Not with me.
Probably because I never got to
fsck.reiserfs --force
while some backup of another partition was present. However I'm sure it
would work as described in the post if I had to run it, because I had
some boot sector backups.
Having backups is important - to the extent that I not only have the
local servers mirrored to backup servers here, but run a nightly rsync
to diskspace on system at my sister's house on the other side of the
country. I do the same for her.
I had looked at journalling file systems in 2001, and decided they were
not mature enough. My logs indicate I didn't switch (to ext3) until
early 2004, when the file-system guru at work approved a switch there.
Having a journaled filesystem can be critical in trying to get systems
back on line. Had a backup generator fail to pick up properly during
an outage, and the UPS had gone about flat - then power failed
completely. We had some servers take 13 hours to fsck when the UPS
didn't have enough left to handle a clean shutdown. In several cases,
it was faster to mkfs and restore from tapes. At least it convinced
management to get a second backup generator, possibly because one of
the file systems that took the longest time to come up was one that
held the ~ filesystems for much of mahogany row (and no, that was not
intentional - fortuitous would be a good word).
As you said it is minor error, if any at all. He didn't referred to
specifications, he simply stated that it has no such signal, and that
is the same as you related not used power good signal to price as
primary design goal.
We looked at the idea of tweaking the 'POWEROK' sensor so that it
triggered much closer to the nominal supply voltage, but the labor
cost was way out of line, and it simply wasn't reliable enough. The
only real way to handle a power outage is a working UPS that can carry
the systems through "short outages" (minutes) and have enough 'poop'
left to do a clean shutdown, and then repeat the entire process to
allow for another failure before the batteries are fully recharged.
Old guy
.
- Follow-Ups:
- Re: [Somewhat OT] Curiouser and curiouser
- From: houghi
- Re: [Somewhat OT] Curiouser and curiouser
- References:
- [Somewhat OT] Curiouser and curiouser
- From: Snorkelson MacBurp
- Re: [Somewhat OT] Curiouser and curiouser
- From: houghi
- Re: [Somewhat OT] Curiouser and curiouser
- From: Rajko M.
- Re: [Somewhat OT] Curiouser and curiouser
- From: Moe Trin
- Re: [Somewhat OT] Curiouser and curiouser
- From: Rajko M.
- [Somewhat OT] Curiouser and curiouser
- Prev by Date: Re: [Somewhat OT] Curiouser and curiouser
- Next by Date: Re: drop outs
- Previous by thread: Re: [Somewhat OT] Curiouser and curiouser
- Next by thread: Re: [Somewhat OT] Curiouser and curiouser
- Index(es):
Relevant Pages
|