tacle_voodoo/howto.txt at master · dalt74/tacle_voodoo · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
0. Build
   # go build -o tacle_voodoo tacle_voodoo.go

1. Retrieve the list of pools with command:
   # ceph osd lspools
   and write pools list in th file pools.cfg

2. Mount all OSD as fuse with ceph-objectstore-tool:
   # ceph-objectstore-tool --op fuse --mountpoint /mnt/osd-0 --data-path /var/lib/ceph/osd/ceph-0/
   # ceph-objectstore-tool --op fuse --mountpoint /mnt/osd-1 --data-path /var/lib/ceph/osd/ceph-1/
   # ceph-objectstore-tool --op fuse --mountpoint /mnt/osd-2 --data-path /var/lib/ceph/osd/ceph-2/

3. Write OSD location into osd.cfg (the sampel file is included too)
   0 filesystem /mnt/osd-0
   1 filesystem /mnt/osd-1
   2 filesystem /mnt/osd-2

4. Scan mounted OSD for the objects.
   # tacle_voodoo scan csv:/mnt/osd-0-objlist 0=/mnt/osd-0
   # tacle_voodoo scan csv:/mnt/osd-1-objlist 1=/mnt/osd-1
   # tacle_voodoo scan csv:/mnt/osd-2-objlist 2=/mnt/osd-2
   You can also scan multiple OSDs into one file:
   # tacle_voodoo scan csv:/mnt/all-objects 0=/mnt/osd-0 1=/mnt/osd-1 2=/mnt/osd-2

4. List volumes found:
   # tacle_voodoo list_volumes <poolname> \
   >    csv:/mnt/osd-0-objlist \
   >    csv:/mnt/osd-1-objlist \
   >    csv:/mnt/osd-2-objlist

5. Request volumes info:
   # tacle_voodoo info <poolname> <rbdname> \
   >    csv:/mnt/osd-0-objlist \
   >    csv:/mnt/osd-1-objlist \
   >    csv:/mnt/osd-2-objlist

6. Extract volume:
   # tacle_voodoo extract <poolname> <rbdname>=<targetfile> \
   >    csv:/mnt/osd-0-objlist \
   >    csv:/mnt/osd-1-objlist \
   >    csv:/mnt/osd-2-objlist

To perfrom all actions except the scan, the tacle_voodoo requires
access to all OSDs mounted as fuse.

If your dead cluster has a lot of OSD and you can't mount'em all
on the single node, you may use tacle_voodoo HTTP data server:

1. Mount some directory on th all OSD hosts via NFS

2. Run tacle_voodoo scan .. on each host and specify store on the NFS mountpoint:
   host-ALPHA # tacle_voodoo scan csv:/nfs/ALPHA-objects 0=/mnt/osd-0 1=/mnt/osd-1
   host-BETA  # tacle_voodoo scan csv:/nfs/BETA-objects 2=/mnt/osd-2 3=/mnt/osd-3

3. Run tacle_voodoo HTTP data server with comand:
   host-ALPHA # tacle_voodoo serve 9000=/mnt/osd-0 9001=/mnt/osd-1
   host-BETA  # tacle_voodoo serve 9002=/mnt/osd-2 9003=/mnt/osd-3
   This will run HTTP gateway to access OSD data over network

4. Specify OSD HTTP servers in osd.cfg - lookup the sample in http_osd.cfg file:
   0 http http://host-ALPHA:9000/
   1 http http://host-ALPHA:9001/
   2 http http://host-BETA:9002/
   3 http http://host-BETA:9003/

5. Run tacle_voodoo for the data access:
   # tacle_voodoo extract \
   > poolname image=recovered-image-filename.raw \
   > csv:/nfs/ALPHA-objects csv:/nfs/BETA-objects
   And the tacle_voodoo will retrieve data via HTTP form the
   multiple hosts based on the URLs in osd.cfg

To use non-default config files (osd.cfg / pools.cfg) sepcify
filenames with env vars OSD_CONFIG and POOL_CONFIG.