[ale] MySQL array based snapshot
DJ-Pfulio
DJPfulio at jdpfu.com
Tue Nov 22 22:35:17 EST 2016
I've seen EMC BCV syncs effectively break I/O on multiple systems. The
local EMC engineer claimed that wasn't possible, but we had outages
during the same period that were eventually traced back to the BCV
snaps. OTOH, gotta wonder who thought daily BCV snaps at 10am was a
good idea. We were doing them multiple times a day, but that was the
only "peak period" time.
OTOH, we did tend to break most HW and most SW at that location.
On 11/22/2016 01:02 PM, Jim Kinney wrote:
> Ah. Vm dude. Ask for his "best method for hot db backup with SAN/NAS
> drive space of vm". Change san to nas as required. I hear a vm
> specialist that's been bit before on performance during backups not
> wanting that pile of meetings ever again.
>
>
> On Nov 22, 2016 11:40 AM, "Chris Fowler" <cfowler at outpostsentinel.com
> <mailto:cfowler at outpostsentinel.com>> wrote:
>
>
>
> ------------------------------------------------------------------------
>
> *From: *"Jim Kinney" <jim.kinney at gmail.com
> <mailto:jim.kinney at gmail.com>>
> *To: *"Atlanta Linux Enthusiasts - Yes! We run Linux!"
> <ale at ale.org <mailto:ale at ale.org>>
> *Sent: *Tuesday, November 22, 2016 10:09:36 AM
> *Subject: *Re: [ale] MySQL array based snapshot
>
>
>
> That's even easier. From inside vm, sync and lock db, vm
> snapshot, unlock. Copy snapshot away. Flatten vm snapshot.
> Vmware and Ovirt/KVM will do this.
>
>
> I've received even more Greek when I asked if they were planning on
> doing snapshots in their storage
>
> "Snaps through VAAI. Our storage is VASA aware. It is the disk
> STUN I am worried about related to running DB."
>
>
> The good news is that the transactions to the DB are not that high.
> It could be if they grow significantly and start monitoring SNMP. I
> store traps in the DB. They could lock the database and that would
> cause all processes to "pause" till it was unlocked. I know this
> because I've received complaints when I've done mysqldumps. Those
> locks are not lock enough to cause timeouts in the apps. They just
> block until the lock is released.
>
> Chris
More information about the Ale
mailing list