Skip to content

Latest commit

 

History

History
75 lines (51 loc) · 2.94 KB

influxdb-zfs.md

File metadata and controls

75 lines (51 loc) · 2.94 KB
Raw

InfluxDB: ZFS Instrumentation

ZFS Monitoring with Telegraf

Telegraf (unit telegraf.service) is used for data acquisition and transfer to InfluxDB.

For monitoring our ZFS servers a custom monitoring script has been written, that collects basic status and performance data and transmits it via the InfluxDB line protocol.

Line Protocol

At this writing the script /opt/sbin/zfsstat-influx is run once per minute by telegraf. Sample output:

zfs,scope=zpool,fset=io,vers=1,ds=zp0 rops=0,wops=670,rbw=0,wbw=53698362
zfs,scope=zpool,fset=space,vers=1,ds=zp0 alloc=7725047697408,free=328278833803264,frag=35,cap=2
zfs,scope=dataset,fset=space,vers=1,ds=zp0 used=24412936676826,avail=214310606658086,compr=1.33
zfs,scope=dataset,fset=space,vers=1,ds=zp0/bcache-zvol used=18920917148922,avail=233231523684704,compr=1.00
zfs,scope=dataset,fset=space,vers=1,ds=zp0/jwd01 used=5464647578208,avail=214310606658086,compr=1.33
zfs,scope=dataset,fset=space,vers=1,ds=zp0/speedtest used=209664,avail=214310606658086,compr=1.00
zfs,scope=dataset,fset=space,vers=1,ds=zp0/upload01 used=209664,avail=214310606658086,compr=1.00

The meaning of the tags is as follows:

  • scope: which type of entity was sampled (currently zpool or dataset)

  • fset: a name for the field set (currently io or space):

    • io: I/O performance data
    • space: data block usage statistics

  • vers: a version number for the fset format (to be incremented whenever the field set layout for a given scope changes)

  • ds: the entity that was sampled (dataset or zpool name)

Observation data:

  • The data of the first field group(s) (rops, wops, rbw, wbw) comes from the command zpool iostat corresponding to the columns "operations read/write" and "bandwidth read/write", the units are ops/s and bytes/s, respectively. This performance data is currently a 5 second snapshot; the sample interval for zpool iostat could be extended if desired, however, in this case the timeout defined in the telegraf config file needs to be changed to match.

    There is one scope=zpool,fset=io field group for each active zpool found on the system.

  • The next field group(s) (alloc, free, frag, cap) report block usage at the zpool level. The data comes from the command zpool list and corresponds to the columns "alloc", "free", "frag" and "cap". The units are bytes (alloc, free) and percentage (fragmentation, pool capacity).

    There is one scope=zpool,fset=space field group for each active zpool found on the system.

  • The remaining field groups (used, avail, compr) report block usage at the dataset ("filesystem") level . The data comes from the command zfs list and reports the ZFS properties "used", "avail" (bytes) and "compressratio" (ratio as floating point).

    There is one scope=dataset,fset=space field group for every dataset or zvol (including the zpool's root data set) found in any of the active zpools on the system.