[Spacewalk-list] Spacewalk in multiple data centers
Grant Gainey
ggainey at redhat.com
Sat Apr 27 13:13:04 UTC 2013
Hi Stefan,
In general, I think you've hit the salient points. Comments inline.
----- Original Message -----
> Hello world,
>
> are there any "best practice" documents for deploying Spacewalk
> servers to multiple data centers in a way that clients will
> load-balance and failover on their own? I think there are a number of
> challenges here:
>
> 1. The actual loadbalancing of HTTP(S) requests. That should be easy
> to handle. Some care should be taken WRT certificates, but that
> shouldn't be a major problem.
Aye, pretty well-understood issues here.
> 2. The (Oracle) database. Should be pretty easy to, set up a physical
> standby and if the primary fails, switchover to the standby
> instance, promote it, edit rhn.conf, restart Spacewalk servers.
Sure, sounds reasonable.
> 3. Actually getting the RPM files from one server to the other -
> rsync'ing them seems like the safest bet.
Don't forget about cobbler's data.
> 4. Taskomatic tasks - this one seems like the only real challenge: We
> don't want both spacewalk servers to e.g. scheduele automatic
> errata updates, or send reports.
Yeah, you only want one taskomatic running, for that very reason. And anything taskomatic does to/from disk (like, say, cobbler-syncing) will only be on disks available to the instance where taskomatic is running.
> Did I miss anything important here? Has anyone already done a setup
> like this and would like to share his/her experience?
If you're looking at redundancy in the same DC, then fast disk shared by all instances keeps your disk-data in sync automatically.
If you actually mean "different data centers", then you'll have to get that dealt with - as you say, rsync or something like it.
I don't know that we have anything formal written up around this - maybe someone else will chime in...
Grant
More information about the Spacewalk-list
mailing list