thank you for response.
yes, I know that we can lost data in this scenario and can not guarantee recover data.But, in my opinion, we need to make Ceph Cluster healthy in spite of data lose.
In this scenario, Ceph cluster has some stuck+stale PGs and goes to Error state.
From the perspective of operation, I would like to say, 'ok, this is disaster situation, and we might forgive damaged RBDs or object data, but we need to make ceph cluster healthy with possible methods that deleting dirty PGs or RBDs, having downtime.. whatever we can.
"healthy" means that Ceph cluster works with no Error State.
I've tried to force delete and re-create some damaged PGs and download rados objects in this situation, but I failed.
and ceph cluster could not response when execute rbd/rados command whith damaged PGs.
I would like to make ceph cluster get back to work.
My approach is not correct?
2017-09-05 10:34 GMT+09:00 David Turner <drakonstein@xxxxxxxxx>:
I don't know that it's still clear what you're asking for. You're understanding that this scenario is going to have lost data that you cannot get back, correct? Some of the information for the RBD might have been in the PGs that you no longer have any copy of. Any RBD that has objects that are now completely gone due to this will be like a harddrive with dead sectors randomly on it. You can't guarantee any of the data on the disk is good, but you might be able to pull some data off of it after the cluster is back to "health_ok" with data loss.Until you recover the cluster to a "healthy" state, stop worrying that you have downtime... you have critical hardware failure and data loss. If this was in production, you should be assessing how many RBD's have lost data, how you're going to try and recover what data you can/need to, and figure out where to go from there... downtime is likely the least of your worries at this point. Being up is useless because the data can't be trusted or used.On Mon, Sep 4, 2017 at 9:23 PM Hyun Ha <hfamily15@xxxxxxxxx> wrote:Hi,I'm still having trouble with above issue.Is anybody there who have same issue or resolve this?Thanks.2017-08-21 22:51 GMT+09:00 Hyun Ha <hfamily15@xxxxxxxxx>:Thanks for response.I can understand why size of 2 and min_size 1 is not an acceptable in production.but, I just want to make the situation of data loss and to know the health of ceph cluster can be clean in this situation(except data recovery because data is gone).So, I've tried to delete PGs using command like "ceph pg 2.67 mark_unfound_lost delete"The result was "Error ENOENT: i don't have pgid 2.67" .So, I confused because I couldn't find the method to delete pg in this situation and to make health_ok of ceph cluster.The only method that I found to make health_ok is this:(when the status of pgs in "stale+active+clean", primary/secondary osd osd.2, osd.6 is gone at the same time.)1. "ceph pg 2.67 mark_unfound_lost delete" doesn't work2. ceph osd crush rm osd.2, osd.63. ceph osd rm osd.2, osd.64. ceph osd auth del osd.2, osd.65. ceph osd lost 2, 6 --yes-i-really-mean-it6. ceph pg force_create_pg 2.67 (but, in this time the status of pg 2.67 is creating..forever)7. re-deploy osd.2, osd.68. stop coresponding osd(primary/secondary osd sequentially for pg 2.67) and wait to complete recovery and start- in this time the primary/secondary osd for pg is not osd.2, osd.6 because pg map is re-created. so I found what is the primary/secondary osd for pg 2.6710. creation of pg 2.67 is done(creating -> peering -> remapped -> active+clean) and ceph become health_okceph cluster become health_ok eventually, but in this time there was a problem that rbd can not found rbd images like below.# rbd ls -p volumeshhvol01# rbd info volumes/hhvol01rbd: error opening image hhvol01: (2) No such file or directorysomething wrong with rbd image but I cannot understand why this happening is occurred.shortly, my question is that:1. Is there any CLI to delete stuck PGs?2. what is correct steps to make health_ok in this situation?Thanks.2017-08-21 20:37 GMT+09:00 David Turner <drakonstein@xxxxxxxxx>:With the exception of trying to re-add the drive to be able to read the data off of it, your only other option is to accept that you lost data and mark the pg as lost and delete it. Not surprisingly, you can't recover the data without any copies of it. Size of 2 is not an acceptable production seeing if data integrity is a priority. Min_size of 1 is even worse. There is plenty of talk about why in the ML archives.
So if you're hoping to not lose data, then your only option is to try and read the data off of the removed osds. If your goal is health_ok regardless of data integrity, then your option is to delete the PGs.
On Mon, Aug 21, 2017, 1:07 AM Hyun Ha <hfamily15@xxxxxxxxx> wrote:Hi, Thank you for response.Details of my pool is below:pool 2 'volumes' replicated size 2 min_size 1 crush_ruleset 0 object_hash rjenkins pg_num 128 pgp_num 128 last_change 627 flags hashpspool stripe_width 0removed_snaps [1~3]My test case was about a scenario of disaster. I think that the situation that all copies of data is deleted can be occurred in production (in my test, I deleted all copy of data by myself because simulate disaster state).When all copy of data is deleted, ceph cluster never get back to clean.How can I recover in this situation?Thank you.2017-08-18 21:28 GMT+09:00 David Turner <drakonstein@xxxxxxxxx>:What were the settings for your pool? What was the size? It looks like the size was 2 and that the PGs only existed on osds 2 and 6. If that's the case, it's like having a 4 disk raid 1+0, removing 2 disks of the same mirror, and complaining that the other mirror didn't pick up the data... Don't delete all copies of your data. If your replica size is 2, you cannot loose 2 disks at the same time.
On Fri, Aug 18, 2017, 1:28 AM Hyun Ha <hfamily15@xxxxxxxxx> wrote:______________________________Hi, Cephers!I'm currently testing the situation of double failure for ceph cluster.But, I faced that pgs are in stale state forever.reproduce steps)0. ceph version : jewel 10.2.3 (ecc23778eb545d8dd55e2e4735b53c c93f92e65b) 1. Pool create : exp-volumes (size = 2, min_size = 1)2. rbd create : testvol013. rbd map and create mkfs.xfs4. mount and create file5. list rados object6. check osd map of each object# ceph osd map exp-volumes rbd_data.4a41f238e1f29.000000000000017a osdmap e199 pool 'exp-volumes' (2) object 'rbd_data.4a41f238e1f29.000000000000017a' -> pg 2.3f04d6e2 (2.62) -> up ([2,6], p2) acting ([2,6], p2) 7. stop primary osd.2 and secondary osd.6 of above object at the same time8. check ceph statushealth HEALTH_ERR16 pgs are stuck inactive for more than 300 seconds16 pgs stale16 pgs stuck stalemonmap e11: 3 mons at {10.105.176.85=10.105.176.85:6789/0,10.110.248.154=10.110. }248.154:6789/0,10.110.249.153= 10.110.249.153:6789/0 election epoch 84, quorum 0,1,2 10.105.176.85,10.110.248.154,10.110.249.153 osdmap e248: 6 osds: 4 up, 4 in; 16 remapped pgsflags sortbitwise,require_jewel_osdspgmap v112095: 128 pgs, 1 pools, 14659 kB data, 17 objects165 MB used, 159 GB / 160 GB avail112 active+clean16 stale+active+clean# ceph health detailHEALTH_ERR 16 pgs are stuck inactive for more than 300 seconds; 16 pgs stale; 16 pgs stuck stalepg 2.67 is stuck stale for 689.171742, current state stale+active+clean, last acting [2,6]pg 2.5a is stuck stale for 689.171748, current state stale+active+clean, last acting [6,2]pg 2.52 is stuck stale for 689.171753, current state stale+active+clean, last acting [2,6]pg 2.4d is stuck stale for 689.171757, current state stale+active+clean, last acting [2,6]pg 2.56 is stuck stale for 689.171755, current state stale+active+clean, last acting [6,2]pg 2.d is stuck stale for 689.171811, current state stale+active+clean, last acting [6,2]pg 2.79 is stuck stale for 689.171808, current state stale+active+clean, last acting [2,6]pg 2.1f is stuck stale for 689.171782, current state stale+active+clean, last acting [6,2]pg 2.76 is stuck stale for 689.171809, current state stale+active+clean, last acting [6,2]pg 2.17 is stuck stale for 689.171794, current state stale+active+clean, last acting [6,2]pg 2.63 is stuck stale for 689.171794, current state stale+active+clean, last acting [2,6]pg 2.77 is stuck stale for 689.171816, current state stale+active+clean, last acting [2,6]pg 2.1b is stuck stale for 689.171793, current state stale+active+clean, last acting [6,2]pg 2.62 is stuck stale for 689.171765, current state stale+active+clean, last acting [2,6]pg 2.30 is stuck stale for 689.171799, current state stale+active+clean, last acting [2,6]pg 2.19 is stuck stale for 689.171798, current state stale+active+clean, last acting [6,2]# ceph pg dump_stuck staleokpg_stat state up up_primary acting acting_primary2.67 stale+active+clean [2,6] 2 [2,6] 22.5a stale+active+clean [6,2] 6 [6,2] 62.52 stale+active+clean [2,6] 2 [2,6] 22.4d stale+active+clean [2,6] 2 [2,6] 22.56 stale+active+clean [6,2] 6 [6,2] 62.d stale+active+clean [6,2] 6 [6,2] 62.79 stale+active+clean [2,6] 2 [2,6] 22.1f stale+active+clean [6,2] 6 [6,2] 62.76 stale+active+clean [6,2] 6 [6,2] 62.17 stale+active+clean [6,2] 6 [6,2] 62.63 stale+active+clean [2,6] 2 [2,6] 22.77 stale+active+clean [2,6] 2 [2,6] 22.1b stale+active+clean [6,2] 6 [6,2] 62.62 stale+active+clean [2,6] 2 [2,6] 22.30 stale+active+clean [2,6] 2 [2,6] 22.19 stale+active+clean [6,2] 6 [6,2] 6# ceph pg 2.62 queryError ENOENT: i don't have pgid 2.62# rados ls -p exp-volumesrbd_data.4a41f238e1f29.000000000000003f ^C --> hangI understand that this is a natural result becasue above pgs have no primary and seconary osd. But this situation can be occurred so, I want to recover ceph cluster and rbd images.Firstly I want to know how to make ceph cluster's state clean.I read document and try to solve this but nothing can help including below commands.- ceph pg force_create_pg 2.6- ceph osd lost 2 --yes-i-really-mean-it- ceph osd lost 6 --yes-i-really-mean-it- ceph osd crush rm osd.2- ceph osd crush rm osd.6- cpeh osd rm osd.2- ceph osd rm osd.6Is there any command to force delete pgs or make ceph cluster clean ?Thank you in advance._________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph. com
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
