Project

General

Profile

Bug #22482

qa: MDS can apparently journal new file on "full" metadata pool

Added by Patrick Donnelly about 5 years ago. Updated almost 5 years ago.

Status:
Won't Fix
Priority:
High
Category:
-
Target version:
-
% Done:

0%

Source:
Development
Tags:
Backport:
luminous
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
Labels (FS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

2017-12-19T06:19:08.951 INFO:teuthology.orchestra.run.smithi202:Running: 'cd /home/ubuntu/cephtest/mnt.0 && sudo dd if=/dev/urandom of=large_file_b bs=1M conv=fdatasync count=50 seek=280'
2017-12-19T06:19:09.086 INFO:teuthology.orchestra.run.smithi202.stderr:dd: error writing 'large_file_b': No space left on device
2017-12-19T06:19:09.086 INFO:teuthology.orchestra.run.smithi202.stderr:1+0 records in
2017-12-19T06:19:09.086 INFO:teuthology.orchestra.run.smithi202.stderr:0+0 records out
2017-12-19T06:19:09.087 INFO:teuthology.orchestra.run.smithi202.stderr:0 bytes copied, 0.0103972 s, 0.0 kB/s
2017-12-19T06:19:09.087 INFO:teuthology.orchestra.run.smithi005:Running: 'sudo adjust-ulimits ceph-coverage /home/ubuntu/cephtest/archive/coverage timeout 120 ceph --cluster ceph osd dump --format=json-pretty'
...
2017-12-19T06:19:09.375 INFO:teuthology.orchestra.run.smithi005.stdout:        {
2017-12-19T06:19:09.375 INFO:teuthology.orchestra.run.smithi005.stdout:            "pool": 6,
2017-12-19T06:19:09.375 INFO:teuthology.orchestra.run.smithi005.stdout:            "pool_name": "cephfs_metadata",
2017-12-19T06:19:09.375 INFO:teuthology.orchestra.run.smithi005.stdout:            "flags": 3,
2017-12-19T06:19:09.375 INFO:teuthology.orchestra.run.smithi005.stdout:            "flags_names": "hashpspool,full",
...
2017-12-19T06:19:09.559 INFO:teuthology.orchestra.run.smithi202:Running: 'cd /home/ubuntu/cephtest/mnt.0 && sudo dd if=/dev/urandom of=small_file_1 bs=1M conv=fdatasync count=0 seek=0'
2017-12-19T06:19:09.659 INFO:teuthology.orchestra.run.smithi202.stderr:0+0 records in
2017-12-19T06:19:09.660 INFO:teuthology.orchestra.run.smithi202.stderr:0+0 records out
2017-12-19T06:19:09.660 INFO:teuthology.orchestra.run.smithi202.stderr:0 bytes copied, 0.000493095 s, 0.0 kB/s
2017-12-19T06:19:09.667 INFO:tasks.cephfs_test_runner:test_full_different_file (tasks.cephfs.test_full.TestClusterFull) ... FAIL

From: /ceph/teuthology-archive/pdonnell-2017-12-19_06:00:26-fs-mimic-dev1-testing-basic-smithi/1979446/teuthology.log

2017-12-19 06:19:09.651 7f5c11f64700  5 mds.0.log _submit_thread 4231952~1606 : EUpdate openc [metablob 0x1, 1 dirs]
2017-12-19 06:19:10.663 7f5c18771700  1 -- 172.21.15.5:6818/1076285882 <== client.4274 172.21.15.202:0/3127382166 36 ==== client_session(???) v1 ==== 28+0+0 (2172597671 0 0) 0x55d3ab46d680 con 0x55d3ab304a00
2017-12-19 06:19:10.663 7f5c18771700 20 mds.0.server get_session have 0x55d3ab246380 client.4274 172.21.15.202:0/3127382166 state open
2017-12-19 06:19:10.663 7f5c18771700  3 mds.0.server handle_client_session client_session(???) v1 from client.4274
2017-12-19 06:19:10.663 7f5c18771700  1 -- 172.21.15.5:6818/1076285882 --> 172.21.15.202:6800/21923 -- osd_op(unknown.0.24:62 6.4 6:2e2fa760:::200.00000001:head [write 37648~1626 [fadvise_dontneed]] snapc 0=[] ondisk+write+known_if_redirected+full_force e42) v9 -- 0x55d3ab3caa40 con 0
2017-12-19 06:19:10.667 7f5c1a712700  1 -- 172.21.15.5:6818/1076285882 <== osd.7 172.21.15.202:6800/21923 41 ==== osd_op_reply(62 200.00000001 [write 37648~1626 [fadvise_dontneed]] v42'33 uv33 ondisk = 0) v9 ==== 157+0+0 (885685870 0 0) 0x55d3ab3caa40 con 0x55d3ab295800

From: /ceph/teuthology-archive/pdonnell-2017-12-19_06:00:26-fs-mimic-dev1-testing-basic-smithi/1979446/remote/smithi005/log/ceph-mds.b.log.gz


Related issues

Related to CephFS - Bug #22483: mds: check for CEPH_OSDMAP_FULL is now wrong; cluster full flag is obsolete Resolved 12/19/2017

History

#1 Updated by Patrick Donnelly about 5 years ago

  • Description updated (diff)

#2 Updated by Patrick Donnelly about 5 years ago

  • Related to Bug #22483: mds: check for CEPH_OSDMAP_FULL is now wrong; cluster full flag is obsolete added

#3 Updated by Patrick Donnelly almost 5 years ago

  • Status changed from New to Won't Fix

This is expected. The MDS is treated specially by the OSDs to allow some writes when the pool is full.

Also available in: Atom PDF