Custom Query (1145 matches)
Results (67 - 69 of 1145)
| Ticket | Resolution | Summary | Owner | Reporter |
|---|---|---|---|---|
| #407 | fixed | athena-auto-update should dpkg --configure -a | jdreed | geofft |
| Description |
A handful of machines in clusters have stopped updating with E: dpkg was interrupted, you must manually run 'sudo dpkg --configure -a' to correct the problem. E: Couldn't rebuild package cache update.log is missing a record of the last upgrade finishing, indicating the system was probably rebooted. I reported this as a theoretical bug five months ago in comment:ticket:217:1 after running into it on my -cluster laptop; as I predicted, it's happening in practice. |
|||
| #412 | duplicate | athena-auto-update should not run when the user is logged in | jdreed | |
| Description |
We decided we don't care about machines being unavailable while they're taking an update. We should change athena-auto-update to run when idle. The desync period should be 2 hours during "low usage" times (2am -8 am) and should be 6 hours otherwise, to help protect against every machine in a cluster taking texlive-full simultaneously. When a machine is being updated, we will touch /etc/nologin and switch to a new VT and display a message to users. We should ensure updates run even on machines that get little-to-no idle time, such as quickstations. |
|||
| #784 | wontfix | athena-auto-update should use start-stop-daemon | jdreed | |
| Description |
It's been noted athena-auto-update should use start-stop-daemon, which avoids the at(1) suckage, avoids frequent cron jobs, and avoids the ConsoleKit? issues with sleeping inside cron jobs. It was noted on the zephyr that we have no idea how start-stop-daemon will behave when the machine is suspend. Because this is a huge re-architecting of auto-update, extensive testing (above and beyond normal test plans) should be conducted. The idea of double-forking was also mentioned. |
|||
