15 minute snapshots? about sanoid HOT 4 CLOSED

Syncoid can do replication whenever you ask it to, and 15 minutes is not
going to be too much overhead for the vast majority of environments.

Sanoid isn't currently set up to manage a 15 minute interval, but I'm
willing to write the code if you want to commission that as a new
feature. Wouldn't be too expensive. Once you commission it and I write
it, it would go into the public codebase, and be maintained along with
the rest of the code.

On 03/18/2016 01:50 PM, Anderath wrote:

It's not clear to me if I can do 15 minute snapshots with this or not.

My thinking is this:

I want to be able to provide a "psuedo-HA" environment where 2 hosts
have enough storage for 2 different sets of VMs to replicate to each
other as a "backup + DR" recovery solution.

So to illustrate:

Host A --Replicates VMs to Host B in the event this host goes down
VM1
VM2
VM3(Cold storage but ready for activation)
VM4(Cold storage but ready for activation)

Host B --Replicates VMs to Host A in the event this host goes down
VM3
VM4
VM1(Cold storage but ready for activation)
VM2(Cold storage but ready for activation)

In this scenario, I believe hourly snapshots may be going too far back
when replicating with a cron running every hour to make sure each host
holds each other hosts' VMs.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
#27

from sanoid.

I hereby commission this as a new feature(Or enhancement... either/or)!

To be honest, I'm fairly new to the open source world(Or github in general). Is there anything apart from testing it that I would need to do? If not, code on and I will vet the code! I've set this up on my work laptop running through a tinc VPN using syncoid every hour. I was thinking about doing some craft pipes through notify-send so I know when a replication fails or the snapshots aren't syncing up for some reason so Sanoid can be run more on a desktop rather than strictly server to server.

I imagine since Sanoid runs every minute, it'll just look at the config file every quarter hour and scoop up another snapshot. With the beautfy of copy on write, it'll just leave down a marker at that point and, wallah, I have my psuedo poor-man's HA failover. As long as 15 minutes in a business day is an acceptable loss(Plus the boot time of whatever VM you're running) then this would be a fairly optimal solution based around ZFS(Even more so, might I say). 👍

from sanoid.

OK, this is awkward.

"Commission" means "offer to pay for development of". I'm probably not
going to spontaneously write 15-minute interval code anytime soon. I'm
willing to review your code if you write it and merge it if it meets
standards, or write it myself if paid to. Sorry!

Another option would be to leave sanoid the way it is, and simply
syncoid more frequently. You'd lose the last 15-minute snapshot every
time you successfully replicated, but your "most recent snapshot" would
always be the syncoid snapshot from the last replication, which would be
inside the window you're looking for.

*/4 * * * * syncoid -r data/images root@otherbox:data/images

That line in your crontab would get you your normal list of
sanoid-maintained snapshots as per policy, plus your most recent
snapshot being the syncoid snapshot from your most recent replication,
with a new replication happening at x:0, x:15, x:30, and x:45.

On 03/22/2016 10:42 PM, Anderath wrote:

I hereby commission this as a new feature(Or enhancement... either/or)!

To be honest, I'm fairly new to the open source world(Or github in
general). Is there anything apart from testing it that I would need to
do? If not, code on and I will vet the code! I've set this up on my
work laptop running through a tinc VPN using syncoid every hour. I was
thinking about doing some craft pipes through notify-send so I know
when a replication fails or the snapshots aren't syncing up for some
reason so Sanoid can be run more on a desktop rather than strictly
server to server.

I imagine since Sanoid runs every minute, it'll just look at the
config file every quarter hour and scoop up another snapshot. With the
beautfy of copy on write, it'll just leave down a marker at that point
and, wallah, I have my psuedo poor-man's HA failover. As long as 15
minutes in a business day is an acceptable loss(Plus the boot time of
whatever VM you're running) then this would be a fairly optimal
solution based around ZFS.

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#27 (comment)

from sanoid.

no further activity since March, so I'm closing this one #wontfix. If anybody wants to commission it, let me know and we can reopen. Also open to pull requests implementing this, but until then, closing.

from sanoid.