mirror of
https://github.com/schoebel/mars
synced 2024-12-28 18:03:12 +00:00
461ac8b4cd
Apparently, sysadmins often forget to execute "marsadm up mydata" (or similar) after a failover. Recall the failover command sequence: "marsadm pause-fetch mydata; marsadm primary --force mydata" Some months later, other sysadmins in the group are stumbling over the very old "pause-fetch" after a regular planned handover via "marsadm primary mydata". It works, but the former primary (which is now secondary) does no longer fetch data, because of the very old pause-fetch command which was never reverted. Afterwards, /mars is filling up slowly over a long time. Somewhen later (e.g. a few days), a monitoring alert "/mars too full" is happening at midnight, leading to an unnecessary on-duty call. A different type of monitoring could help, by not only tracking the filling level of /mars, but also view-todo-fetch or similar. However, some people dislike this, because there exist operational use cases (like creation of backups) where pause-fetch is executed _deliberately_ for a longer time. Here is a workaround for a forgotten resume-fetch / up after the first failover: After the _original_ "marsadm primary" or "primary --force" has succeeded by appearance of /dev/mars/mydata, we simply execute the equivalent of "marsadm up mydata". This changes the semantics of the "primary" command. Hopefully no scripts on this world will break. |
||
|---|---|---|
| .. | ||
| cron.d | ||
| udev | ||
| make-man.sh | ||
| mars-log-impex.c | ||
| marsadm | ||
| marsadm.8 | ||
| write-reboot.c | ||