After talking with Dan and Max yesterday I thought I'd send out how serial console currently works on consolo. I thought I had done this before but just in case, here it is. I've configured the two systems' Opengears to accept both SSH and RFC2217 connections. The RFC2217 should be much lighter weight on the OGs though both OGs have been up over 90 days (which is the last time I configured them to use SSH) so I'm not sure the SSH overhead/reboot problem is still there in their current firmware (3.7.0u1). In any case the configuration on the OGs is pretty current: each port is labeled in the UI to what's on that port and that label matches the definition in conserver. The OGs do RADIUS first (through the hubs so Yubi if you're configured, CC otherwise) and then local (root user password is MCS root). Conserver on consolo is now configured to talk to the two systems' OGs via RFC2217 and should keep persistent connections so that logs in /var/log/conserver tail the console. Conserver talks to the bio OGs still via SSH and I can help Dan to configure them to do RFC2217 if he wants. Conserver also has consoles via remote virsh to all the VMs on the vservers. consolo must be able to ssh as root to all the vservers in order to do this. These connections are also persistent so if you need console on a VM you should go through consolo and not via virsh on the vserver itself as weird things happen when two processes try to access the virsh console. These consoles are also logged to /var/log/conserver for forensics of what happened on a VM console after the fact. The conserver.cf file should be automatically created via daily cronjob on enterprise - /etc/cron.daily/create-vm-conserver-entries.sh - using the automatically mined information from the VM listing on enterprise - /tmp/vm-list. It does this by copying a header file on consolo - /etc/conserver/conserver.cf_header - to /etc/conserver/conserver.cf and then creating the VM entries, finally restarting the server. So if you need console for any machine, you should go through consolo. If you need to configure a new VM for console access, do nothing, it should be created daily. If you need to configure a new physical server for console edit the header file on consolo not the actual conserver.cf. Let me know if none of this makes sense. Also we should discuss merging consolo and enterprise to reduce the number of machines that have unfettered root.
I also put this on the wiki: https://sharepoint.anl.gov/Divisions/mcs/operations/Systems%20Documentation%... On Aug 13, 2013, at 8:38 AM, Ti Leggett <[email protected]> wrote:
After talking with Dan and Max yesterday I thought I'd send out how serial console currently works on consolo. I thought I had done this before but just in case, here it is.
I've configured the two systems' Opengears to accept both SSH and RFC2217 connections. The RFC2217 should be much lighter weight on the OGs though both OGs have been up over 90 days (which is the last time I configured them to use SSH) so I'm not sure the SSH overhead/reboot problem is still there in their current firmware (3.7.0u1). In any case the configuration on the OGs is pretty current: each port is labeled in the UI to what's on that port and that label matches the definition in conserver. The OGs do RADIUS first (through the hubs so Yubi if you're configured, CC otherwise) and then local (root user password is MCS root).
Conserver on consolo is now configured to talk to the two systems' OGs via RFC2217 and should keep persistent connections so that logs in /var/log/conserver tail the console. Conserver talks to the bio OGs still via SSH and I can help Dan to configure them to do RFC2217 if he wants.
Conserver also has consoles via remote virsh to all the VMs on the vservers. consolo must be able to ssh as root to all the vservers in order to do this. These connections are also persistent so if you need console on a VM you should go through consolo and not via virsh on the vserver itself as weird things happen when two processes try to access the virsh console. These consoles are also logged to /var/log/conserver for forensics of what happened on a VM console after the fact.
The conserver.cf file should be automatically created via daily cronjob on enterprise - /etc/cron.daily/create-vm-conserver-entries.sh - using the automatically mined information from the VM listing on enterprise - /tmp/vm-list. It does this by copying a header file on consolo - /etc/conserver/conserver.cf_header - to /etc/conserver/conserver.cf and then creating the VM entries, finally restarting the server.
So if you need console for any machine, you should go through consolo. If you need to configure a new VM for console access, do nothing, it should be created daily. If you need to configure a new physical server for console edit the header file on consolo not the actual conserver.cf.
Let me know if none of this makes sense. Also we should discuss merging consolo and enterprise to reduce the number of machines that have unfettered root.
participants (1)
-
Ti Leggett