What is Scrubbing?¶
Ceph scrubbing is analogous to
fsckon the object storage layer.
-- Taken from Ceph OSD Scrubbing config reference documentation.
There's two scrubbing "modes":
- ("Light") Scrubbing - Verify basic object attributes and size.
- "Deep" Scrubbing - Read object data and verify checksums.
Especially deep scrubbing can cause performance degredation when it is active at the "wrong time" (e.g., your applications peak hours).
Should I disable Scrubbing?¶
No, don't ever disable scrubbing (forever) in a Ceph cluster! It is vital to have Ceph verify data integrity automatically and regurarly so it can detect and repair any errors found as early as possible.
My cluster is reporting
inconsistent PGs. What should I do?¶
Before attempting any manual intervention, please read up on the
ceph pg repair command here: More Information on PG Repair - Ceph Documentation.
After you have made sure you understand what has probably happen to (data) objects in your PG(s), you can attempt to repair them by following the steps here:
The default value is
3, you can try reducing it to
2 or even to
1, in case where you want scrubbing to cause lower impact when scrubbing happens.
The maximum interval in seconds for scrubbing the Ceph OSD Daemon irrespective of cluster load.
If this threshold is crossed, scrubbing can and will happen no matter the current cluster load, unless disabled using
To have the scrubbing happen mainly during 18:00 to 05:00 on the weekend this config can help you:
Please keep in mind that scrubbing might continue after the schedule when the
osd_scrub_max_interval is reached.
Weekday evening/night scrubbing¶
Only ever deactivate scrubbing if really necessary or for short periods of time.
Man page of
ceph command, check out the
ceph set subcommand section.
Be sure to check out Ceph's documentation on OSD scrubbing as it contains all config options and their values.