Project

General

Profile

Actions
Copy link

Bug #23251

closed

ceph daemon osd.NNN slow_used_bytes and slow_total_bytes wrong?

Added by Ben England about 6 years ago. Updated about 6 years ago.

Status:
Rejected
Priority:
Normal
Assignee:
-
Target version:
Ceph - v12.2.4
% Done:

0%

Source:
other
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
Ceph - v12.2.1
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

version: ceph-osd-12.2.1-34.el7cp.x86_64 = RHCS 3.0z1

In trying to understand ceph daemon osd.NNN perf dump counters, I came across this riddle. Bluestore is supposed to return the total amount of space available on the "slow" (data) device and the amount in use with:

[root@b10-h01-r620 bene]# ssh c05-h21-6048r ceph daemon osd.240 perf dump | grep slow
"slow_total_bytes": 79989571584, (~80 GB)
"slow_used_bytes": 3394240512, (~3 GB)

However, if I look at OSD-level counters:

[root@b10-h01-r620 bene]# ssh c05-h21-6048r ceph daemon osd.240 perf dump | grep stat_bytes
"stat_bytes": 2000811954176, (~2 TB)
"stat_bytes_used": 1252726378496, (~1.25 TB)
"stat_bytes_avail": 748085575680,

stat_bytes matches what parted says about the partition size, so I believe this one and not the bluestore one. But how does the allocator know if it has free space when slow_total_bytes and slow_used_bytes are wrong?

Anyone else see this on newer versions?

Actions
Copy link

Also available in: Atom PDF