Re: [luv-main] backkuppc and ZFS and now memlockd too (was Re: Back ups)
Quoting "Chris Samuel" <chris@csamuel.org>:
On Mon, 12 Sep 2011 10:32:43 AM Peter Ross wrote:
Quoting "Andrew McGlashan" <andrew.mcglashan@affinityvision.com.au>:
including a working fsck ability, which is essential.
It's overrated;-)
Not on an experimental filesystem which can have some, umm, interesting bugs at times.. ;-)
The fsck looks more magic than everything else to me (because it has to catch "the unexpected") and if you try "magic" on something that is not fully understood yet.. good luck with that. I see your point. At the moment I go for plan B first (backup/failover) instead of praying that magic helps me if it goes pear-shaped. In that sense it may be a waste of time to write a fsck. However, I can imagine that it helps you to improve the original code. If you write fsck you have to think about "what can go wrong" - and then you are half-way through to fix it (and not relying on fsck). I always avoid software that needs "repair tools" in production. E.g. I hated to work with MySQL 3 and having to fix MyISAM tables and indexes if something went wrong. It took me a while until I wanted to touch MySQL again.. These days I do not even remember what the repair tools look like, and I hope I never have to google for it. Regards Peter
Hi Peter, Peter Ross wrote:
The fsck looks more magic than everything else to me (because it has to catch "the unexpected") and if you try "magic" on something that is not fully understood yet.. good luck with that.
Yes, I do agree with that.
At the moment I go for plan B first (backup/failover) instead of praying that magic helps me if it goes pear-shaped.
Correct, but we need to be able to move data faster b/w disks -- restores can take too long with most of the disks we have today, bring on the faster storage ASAP.... do I hear "racetrack ;-)" or fast and cheap SSDs. Yeah, back on that point, we'll have 10GE cards one day as normal instead of just GE -- another one for Russell to pick over ;-)
In that sense it may be a waste of time to write a fsck. However, I can imagine that it helps you to improve the original code. If you write fsck you have to think about "what can go wrong" - and then you are half-way through to fix it (and not relying on fsck).
Yes, absolutely.
I always avoid software that needs "repair tools" in production. E.g. I hated to work with MySQL 3 and having to fix MyISAM tables and indexes if something went wrong.
Fortunately I never used MySQL that long ago .... I use it now, but I also use Oracle plenty. I think that Postgresql is the way to go longer term though, but I haven't even done anything with it yet.
It took me a while until I wanted to touch MySQL again.. These days I do not even remember what the repair tools look like, and I hope I never have to google for it.
;-) At least we have Google and many, many reference points these days when things do go wrong and we need them. You don't always get the answers, but it is far easier today to find the same problems experienced by others as so many people are using the tools. Cheers -- Kind Regards AndrewM Andrew McGlashan
On Wed, 14 Sep 2011, Andrew McGlashan <andrew.mcglashan@affinityvision.com.au> wrote:
Correct, but we need to be able to move data faster b/w disks -- restores can take too long with most of the disks we have today, bring on the faster storage ASAP.... do I hear "racetrack ;-)" or fast and cheap SSDs. Yeah, back on that point, we'll have 10GE cards one day as normal instead of just GE -- another one for Russell to pick over ;-)
It seems to me that the main benefit that most home users received from GigE over 100baseT was the fact that crossover cables are no longer needed. Most corporations don't receive any benefit, even bloated MS-Office documents can be loaded fast enough on 100baseT. Sales of GigE Ethernet cards never became popular. GigE became commonly used on the desktop when it became a standard motherboard feature. I expect that when 10GigE becomes popular on the desktop it will be because it's integrated into motherboard chipsets. So it won't be an issue of buying a SATA3 capable motherboard now to support 10GigE because people who want it will just buy a motherboard that has it. If you want faster restores then just use the fastest 1/3 of each disk. With 3TB disks that's quite a viable option. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/
participants (3)
-
Andrew McGlashan -
Peter Ross -
Russell Coker