Second Satellite for testing purposes
Hi all,
we have RH Satellite installed (and of course licensed). I have just seen that a new 5.6 has just been released. I don't want to test it on production and ideally it would be good to test upgrade process and functionality on another, testing server. How to cope with it when we have one satellite server licence bought ?
How do you guys cope with ? Do you have several licences ?
Responses
We have Satellite server as virtual guest. Before upgrade I shutdown, take snapshot, boot, upgrade and test. Easy rollback if something goes wrong.
Hardware system I would take it down and make offline full backup with rescue CD. Time consuming, i know :-)
How is your phyisical server's storage setup?
- If you're using LVM or equivalent technologies (even array-based), you could always leverage the storage-snapshotting capability to create a rollback point.
- If your physical server uses locally-mirrored drives, you could always pop one of the drives out as a state-save (and put it on a shelf, somewhere) then do your testing.
If you have a virtualization infrastructure available, you could always do a clone-P2V your production Satellite server, then use that for your testing.
Downside of either of the bulleted options is, from a client perspective, it's still a live-fire type of exercise. The virtualization option is still likely to be your overall "safest" path.
Red Hat does offer the ability to acquire a short-term Satellite Subscription. I believe mine was 90-day and did not include many of the additional channels, etc.. - so, any real testing I did had to take that in to consideration. I believe you can negotiate longer terms, but that doesn't address the real need.
This thread has me wondering if you could build the Satellite (in disconnected mode) and also perform the upgrade (offline as well). Satellite has the ability to have completely disconnected deployments where you have to export the data from one and import to the disconnected system. That might be something to pursue?
We recently upgraded our Red Hat Satellite 5.5 (embedded Oracle) to 5.6 (embedded PostgreSQL) on a running RHEL6.4+. The upgrade process was quite straightforward and Satellite seems to perform better (more interactive) now than before. We have about 2300 servers and about 17 different base/cloned channels with each about 15 subchannels. In total the database was 33GB and the on-disk repositories was 278GB. After the upgrade the vanilla PostgreSQL was only 18GB.
Do make an (offline) backup of the Satellite and (if possible) a snapshot of your VM before starting. Also make sure you have sufficient free disk-space in /var/lib/pgsql (we needed more than 60GB for temporary table spaces after the upgrade).
Do follow the procedure rigidly:
- Red Hat Satellite 5.6 - Installation Guide - Chapter 6. Upgrades
- /etc/sysconfig/rhn/satellite-upgrade/README
It took about 3 hours for the upgrade to complete, and about 24h for the Satellite to settle to normal (idle) I/O behaviour (due to some outstanding scheduled repo/errata tasks).
We differ from most places, we run 2 satellites for redundancy and for testing. So when we did the upgrade we were able to bring one down do the upgrade. Then we could just build the other one off the upgraded one.
Przemyslaw Bak,
Remember that the partitioning is different for satellite version 5.6 (you probably know that). Namely, the database is now under /var/lib/pgsql/ (See Dag Wieer's note on space)
David Powles' idea is a good one and so is Joe's if you can afford to have two persistent satellite servers (one for testing and another).
In our environment, we also have a develoment one - but to tell you the truth, I personally believe the straight install (new server) Satellite version 5.6 on rhel 6.current is rather straightforward and the install was smooth (we have seven of them across two different Red Hat accounts). And so far, we have SELinux running on it with no issue.
For those that may use vmware, for /var/satellite, I have a very large (overkill, but we can afford the space, rhel 5 and 6 channels, using 126g of 316g total) 316GB partition under lvm. However becuase vmware will not (in our past) snapshot a very large lvm, I made three 110gb vmdk files, and in the kickstart, made the system use lvm to tie those three together as one disk and we've been running that with no issues (and snapshots work) on those servers under vmware. (I realize your server is physical, this is for those who might consider vmware)
We do have one physical satellite server and we put RHEL 6.current with Satellite 5.6 and it's installation was smooth.
one additional thing to remember - do not join the server to ldap or another authentication server [UNTIL AFTER install] because if you have an apache user already defined under LDAP, - the satellite install script will NOT create an apache server and when you go to build the bootstrap file, that's the first place you'll see the problem (because the apache user is used there). We got past that problem but wait until the satellite server install is complete before joining to LDAP or removing users that the satellite software will use.
Good luck
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
