Normal rate monitoring should be preformed on parent HTB classes/queues.
SFQ classes are dynamically allocated (as queues fill) and given random names and this messes up charts.
Normal rate monitoring should be preformed on parent HTB classes/queues.
SFQ classes are dynamically allocated (as queues fill) and given random names and this messes up charts.
* redirect error message of `ls` to /dev/null
* fetch mismatch_cnt to get correct count
* https://www.kernel.org/doc/Documentation/md.txt
* sync_completed: the number of sectors that have been completed
* mismatch_cnt: the number of errors that are found
* mdadm needs super-user privileges, but /proc/mdstat is readable for nobody
* fetch an array state from mdstat 'reshape|check|resync|recovery'
* reshape, recovery: rebuilding array -> set percent to _rebuild.value
* check, resync: data scrubbing or mirror rebuilding => set percent to _check.value
* resync=DELAYED|PENDING => set both to zero, because details are unknown even if using mdadm
* more info about /proc/mdstat => kernel src:/drivers/md/md.c, md_seq_show()
* one line regular expression matching to multiline text is a bit hard to picture
* apply regular expressions to each line
* first line should like "active raid1 sda1[0] sdc1[2] sdb1[1]"
* second line should like "123456 blocks super 1.2 [2/2] [UU]"
* third line will appear when the array is in action
Modified version of plugins from https://github.com/pcdummy/mongomon that composes all the plugins into one script.
It works fine, but still leaves a room for future improvement if anyone is interested.
Providing a --upper-limit parameter is only useful if the size of the
filesystem never changes. On ZFS, this assumption is only true for the
zpool; for all other filesystems, the available space is shared, and the
totalled values will vary. When this is the case, the --upper-limit
parameter will obscure peaks in the graph.
- Allow log file path to be configured via plugin-conf.d env.logfile
- Add counters for successful and failed renames
- Switch to use DERIVE and min 0 to fix problems during rotation and show request counts per period
- Improve performance for large log files by using awk
If you have a directory where daily, weekly, monthly you generate
backups of files, SQL databases, etc, you might want to be informed
if the backup generation process is gone or the behavior changes
greatly. This monitoring plugin watches the number of young files
in a certain directory. Relies on find command to do the job.
Newer versions of mtr outputs a "Start: <date here>" line which would be
picked up and incorrectly graphed in munin. Adding an additional inverse grep
field to remove this extra line if it is present. Also added case insensitive
flag to the grep.
