PG (Placement Group) notes
Miscellaneous copy-pastes from emails, when this gets cleaned up it should move out of /dev.
PG = “placement group”. When placing data in the cluster, objects are mapped into PGs, and those PGs are mapped onto OSDs. We use the indirection so that we can group objects, which reduces the amount of per-object metadata we need to keep track of and processes we need to run (it would be prohibitively expensive to track eg the placement history on a per-object basis). Increasing the number of PGs can reduce the variance in per-OSD load across your cluster, but each PG requires a bit more CPU and memory on the OSDs that are storing it. We try and ballpark it at 100 PGs/OSD, although it can vary widely without ill effects depending on your cluster. You hit a bug in how we calculate the initial PG number from a cluster description.
There are a couple of different categories of PGs; the 6 that exist
(in the original emailer’s
ceph -s output) are “local” PGs which
are tied to a specific OSD. However, those aren’t actually used in a
standard Ceph configuration.
Mapping algorithm (simplified)
Many objects map to one PG.
Each object maps to exactly one PG.
One PG maps to a single list of OSDs, where the first one in the list is the primary and the rest are replicas.
Many PGs can map to one OSD.
A PG represents nothing but a grouping of objects; you configure the number of PGs you want, number of OSDs * 100 is a good starting point , and all of your stored objects are pseudo-randomly evenly distributed to the PGs. So a PG explicitly does NOT represent a fixed amount of storage; it represents 1/pg_num’th of the storage you happen to have on your OSDs.
Ignoring the finer points of CRUSH and custom placement, it goes something like this in pseudocode:
locator = object_name obj_hash = hash(locator) pg = obj_hash % num_pg OSDs_for_pg = crush(pg) # returns a list of OSDs primary = osds_for_pg replicas = osds_for_pg[1:]
If you want to understand the crush() part in the above, imagine a perfectly spherical datacenter in a vacuum ;) that is, if all OSDs have weight 1.0, and there is no topology to the data center (all OSDs are on the top level), and you use defaults, etc, it simplifies to consistent hashing; you can think of it as:
def crush(pg): all_osds = ['osd.0', 'osd.1', 'osd.2', ...] result =  # size is the number of copies; primary+replicas while len(result) < size: r = hash(pg) chosen = all_osds[ r % len(all_osds) ] if chosen in result: # OSD can be picked only once continue result.append(chosen) return result
User-visible PG States
diagram of states and how they can overlap
the PG is still being created
requests to the PG will be processed
all objects in the PG are replicated the correct number of times
a replica with necessary data is down, so the pg is offline
recovery could not finish because object(s) are unfound.
backfill could not finish because object(s) are unfound.
the PG is in a quiesced-IO state due to an impending PG merge. That happens when pg_num_pending < pg_num, and applies to the PGs with pg_num_pending <= ps < pg_num as well as the corresponding peer PG that it is merging with.
the PG is being checked for inconsistencies
some objects in the PG are not replicated enough times yet
replicas of the PG are not consistent (e.g. objects are the wrong size, objects are missing from one replica after recovery finished, etc.)
the PG is undergoing the Peering process
the PG is being checked and any inconsistencies found will be repaired (if possible)
objects are being migrated/synchronized with replicas
the PG is waiting in line to start backfill
a pg is missing a necessary period of history from its log. If you see this state, report a bug, and try to start any failed OSDs that may contain the needed information.
the PG is in an unknown state - the monitors have not received an update for it since the PG mapping changed.
the PG is temporarily mapped to a different set of OSDs from what CRUSH specified
In conjunction with scrubbing the scrub is a deep scrub
a special case of recovery, in which the entire contents of the PG are scanned and synchronized, instead of inferring what needs to be transferred from the PG logs of recent operations
backfill reservation rejected, OSD too full
the PG is waiting for the local/remote recovery reservations
the PG can’t select enough OSDs given its size
the PG is peered but not yet active
the PG peered but can’t go active
the PG is trimming snaps
the PG is queued to trim snaps
recovery reservation rejected, OSD too full
the PG could not complete snap trimming due to errors
the PG has been marked for highest priority recovery
the PG has been marked for highest priority backfill
an attempt to repair the PG has failed. Manual intervention is required.
Omap statistics are gathered during deep scrub and displayed in the output of the following commands:
ceph pg dump ceph pg dump all ceph pg dump summary ceph pg dump pgs ceph pg dump pools ceph pg ls
As these statistics are not updated continuously they may be quite inaccurate in an environment where deep scrubs are run infrequently and/or there is a lot of omap activity. As such they should not be relied on for exact accuracy but rather used as a guide. Running a deep scrub and checking these statistics immediately afterwards should give a good indication of current omap usage.