Openstack changes
- Install ceilometer to identify unused instances, contact users to see if instances still needed after identified as unused
- Setup test Mitaka cluster
- Decide whether to pursue dynDNS changes in Liberty, or wait for Mitaka because it's builtin
- Install new storage and compute nodes (in progress)
- Set up shared data store for instances to enable live migrations (in progress — tested and works, but uplink cable on switch needs to be fixed because we're getting 20-60% packet loss on all new nodes)
- Make DXL storage nodes available (in progress)
- Install Swift
Create new storage network (finished 10may16)Increase MTU to 9000 on all networks (finished 10may16)- Investigate Amazon EC2
- Download information about Nebula nightly into database so we can get information about nodes if they have been deleted, etc.
System Administration
- Write script to delete old log files if disk is at 95%
- Set up stunnel from compute nodes to controller for SMTP
Set up stunnel from controller to compute nodes for rsync to grab glance images? (unnecessary now that glance and nova will have a GlusterFS backend)- Write script to monitor nova, cinder services and notify/restart if necessary
- Write scripts to monitor resource usage
- Install larger disk on controller?
- Set up syslog aggregator on controller
- Forward IPMI syslog to controller
- Forward node syslog to controller
- Move /var/spool/mqueue on controller to a new partition; when a service goes haywire and logs fill the disk, we can't receive email warnings
- Move database dumps from ~root/backups to a different space (perhaps GlusterFS?)
Send compute3 in for repair- Move /var on controller to the larger /home partition; should buy time if logs are filling (20x more space there)
- Write scripts to monitor actual disk space for gluster volumes
- Put scripts into repository and check out to /usr/local/scripts on nebula
Create option on node installation script to not run the global commands (if they've already been run, for example) (finished 10may16)