FTP Emergency Docs
This doc will describe a variety of potential emergency situations we may run into with the ftp cluster. Its mostly to be used as a potential guide in case they happen.
FTP Array Failure
In the eventual but hopefully unlikely event that an array of disks completely fails due to too many disks failing, this might be what to do. This is much easier to work around if the machine is one of the slaves (nyc/chi), however in the case of ftp-osl going down it gets even more complicated.
Some things to consider:
- This machine acts as master mirror for some projects
- All slaves sync from this host
- While we rebuild ftp-osl, all new syncs will likely be stopped to ensure data integrity
With this in mind, we have to consider a few list of actions. The TL;DR version of what we should probably do is:
- Stop all cronjobs on all ftp hosts to mitigate any potential data loss (simply stopping cron is probably ok)
- Take ftp-osl out of DNS rotation
- Shutdown http/xinetd. We might want to restrict ssh access too
- Notify the outage via hosting list and other methods (may even need a news post on our website)
- Ensure the integrity of the data on the other two slave machines for the same partition (basically make sure each repo is the same size or close to it)
- Designate which slave to use to re-copy data back to ftp-osl, likely chi but nyc might be faster
- Rebuild the disk array and mkfs.xfs it, etc. It's probably easier to just recreate the array at that point
- Start an rsync from slave node (this will likely take DAYS depending on the load and such).
- Ensure everything looks sane
- Re-enable cronjobs and do some tests
- Add back into rotation
- Send out another announcement