Como você adiciona um HDD de substituição a um volume glusterfs?

Question

Eu recentemente herdei uma configuração do gluster que eu sei literalmente sobre zero. Um dos HDDs que fornece um tijolo para o volume falhou e eu consegui substituir esse HDD e o sistema operacional host pode ver o HDD. Eu consegui formatá-lo com sucesso e ele está na posição em que o HDD substituído está agora montado como o HDD que ele substituiu.

Aqui é onde eu preciso de ajuda.

Acredito que preciso executar algum tipo de comando de recuperação, mas estou bastante confuso sobre como fazer isso com o GlusterFS. Eis algumas das informações de fundo.

$ mount |grep glus
/dev/sdc1 on /data/glusterfs/sdc1 type xfs (rw,relatime,attr2,inode64,noquota)
/dev/sdg1 on /data/glusterfs/sdg1 type xfs (rw,relatime,attr2,inode64,noquota)
/dev/sdf1 on /data/glusterfs/sdf1 type xfs (rw,relatime,attr2,inode64,noquota)
/dev/sdb1 on /data/glusterfs/sdb1 type xfs (rw,relatime,attr2,inode64,noquota)
/dev/sdd1 on /data/glusterfs/sdd1 type xfs (rw,relatime,attr2,inode64,noquota)
127.0.0.1:/nova on /var/lib/nova/instances type fuse.glusterfs (rw,relatime,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072)
127.0.0.1:/cinder on /var/lib/nova/mnt/92ef2ec54fd18595ed18d8e6027a1b3d type fuse.glusterfs (rw,relatime,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072)
/dev/sde1 on /data/glusterfs/sde1 type xfs (rw,relatime,attr2,inode64,noquota)

O disco rígido que eu substituí é /dev/sde1 . Eu tenho montado (como visto acima) e quando eu corro glusterfs volume info vejo que está listado lá:

$ gluster volume info nova

Volume Name: nova
Type: Distributed-Replicate
Volume ID: f0d72d64-288c-4e72-9c53-2d16ce5687ac
Status: Started
Number of Bricks: 10 x 2 = 20
Transport-type: tcp
Bricks:
Brick1: icicle07:/data/glusterfs/sdb1/brick
Brick2: icicle08:/data/glusterfs/sdb1/brick
Brick3: icicle09:/data/glusterfs/sdb1/brick
Brick4: icicle10:/data/glusterfs/sdb1/brick
Brick5: icicle11:/data/glusterfs/sdb1/brick
Brick6: icicle07:/data/glusterfs/sdc1/brick
Brick7: icicle08:/data/glusterfs/sdc1/brick
Brick8: icicle09:/data/glusterfs/sdc1/brick
Brick9: icicle10:/data/glusterfs/sdc1/brick
Brick10: icicle11:/data/glusterfs/sdc1/brick
Brick11: icicle07:/data/glusterfs/sdd1/brick
Brick12: icicle08:/data/glusterfs/sdd1/brick
Brick13: icicle09:/data/glusterfs/sdd1/brick
Brick14: icicle10:/data/glusterfs/sdd1/brick
Brick15: icicle11:/data/glusterfs/sdd1/brick
Brick16: icicle07:/data/glusterfs/sde1/brick
Brick17: icicle08:/data/glusterfs/sde1/brick
Brick18: icicle09:/data/glusterfs/sde1/brick
Brick19: icicle10:/data/glusterfs/sde1/brick
Brick20: icicle11:/data/glusterfs/sde1/brick

Tentar executar um comando de recuperação resulta nisso:

$ gluster volume heal nova full
Locking failed on c551316f-7218-44cf-bb36-befe3d3df34b. Please check log file for details.
Locking failed on 79a6a414-3569-482c-929f-b7c5da16d05e. Please check log file for details.
Locking failed on ae62c691-ae55-4c99-8364-697cb3562668. Please check log file for details.
Locking failed on 5f43c6a4-0ccd-424a-ae56-0492ec64feeb. Please check log file for details.
Locking failed on cb78ba3c-256f-4413-ae7e-aa5c0e9872b5. Please check log file for details.
Locking failed on 6c0111fc-b5e7-4350-8be5-3179a1a5187e. Please check log file for details.
Locking failed on 88fcb687-47aa-4921-b3ab-d6c3b330b32a. Please check log file for details.
Locking failed on d73de03a-0f66-4619-89ef-b73c9bbd800e. Please check log file for details.
Locking failed on c7416c1f-494b-4a95-b48d-6c766c7bce14. Please check log file for details.
Locking failed on 4a780f57-37e4-4f1b-9c34-187a0c7e44bf. Please check log file for details.

As tentativas de executar o comando novamente resultam nisso:

$ gluster volume heal nova full
Another transaction is in progress. Please try again after sometime.

Reiniciar o glusterd irá liberar esse bloqueio, mas não sei o que o comando de recuperação acima está tentando me dizer. Os registros que eu acho inúteis, já que existem vários, e não são totalmente claros para mim, o que combina com o seguinte:

$ ls -ltr /var/log/glusterfs
...
rw------- 1 root root      41711 Aug  1 00:51 glfsheal-nova.log-20150801
-rw------- 1 root root          0 Aug  1 03:39 glfsheal-nova.log
-rw------- 1 root root       4297 Aug  1 14:29 cmd_history.log-20150531
-rw------- 1 root root     830449 Aug  1 17:03 var-lib-nova-instances.log
-rw------- 1 root root     307535 Aug  1 17:03 glustershd.log
-rw------- 1 root root     255801 Aug  1 17:03 nfs.log
-rw------- 1 root root       4544 Aug  1 17:12 cmd_history.log
-rw------- 1 root root      28063 Aug  1 17:12 cli.log
-rw------- 1 root root   17370562 Aug  1 17:14 etc-glusterfs-glusterd.vol.log
-rw------- 1 root root 1759170187 Aug  1 17:14 var-lib-nova-mnt-92ef2ec54fd18595ed18d8e6027a1b3d.log

Qualquer orientação seria apreciada.

EDIT # 1

Parece que o sistema está tendo problema quando tenta exibir o glusterfsd correspondente para o brick / HDD que eu adicionei de volta. Aqui está a saída do arquivo de log: /var/log/glusterfs/bricks/data-glusterfs-sde1-brick.log :

[2015-08-01 21:40:25.143963] I [MSGID: 100030] [glusterfsd.c:2294:main] 0-/usr/sbin/glusterfsd: Started running /usr/sbin/glusterfsd version 3.7.0 (args: /usr/sbin/glusterfsd -s icicle11 --volfile-id nova.icicle11.data-glusterfs-sde1-brick -p /var/lib/glusterd/vols/nova/run/icicle11-data-glusterfs-sde1-brick.pid -S /var/run/gluster/d0a51f364706915faa35c6cca46e9ce6.socket --brick-name /data/glusterfs/sde1/brick -l /var/log/glusterfs/bricks/data-glusterfs-sde1-brick.log --xlator-option *-posix.glusterd-uuid=5e09f3ec-bfbc-490b-bd93-8e083e8ebd05 --brick-port 49155 --xlator-option nova-server.listen-port=49155)
[2015-08-01 21:40:25.190863] I [event-epoll.c:629:event_dispatch_epoll_worker] 0-epoll: Started thread with index 1
[2015-08-01 21:40:48.359478] I [graph.c:269:gf_add_cmdline_options] 0-nova-server: adding option 'listen-port' for volume 'nova-server' with value '49155'
[2015-08-01 21:40:48.359513] I [graph.c:269:gf_add_cmdline_options] 0-nova-posix: adding option 'glusterd-uuid' for volume 'nova-posix' with value '5e09f3ec-bfbc-490b-bd93-8e083e8ebd05'
[2015-08-01 21:40:48.359696] I [server.c:392:_check_for_auth_option] 0-/data/glusterfs/sde1/brick: skip format check for non-addr auth option auth.login./data/glusterfs/sde1/brick.allow
[2015-08-01 21:40:48.359709] I [server.c:392:_check_for_auth_option] 0-/data/glusterfs/sde1/brick: skip format check for non-addr auth option auth.login.a9c47852-7dcf-4f89-80e5-110101943f36.password
[2015-08-01 21:40:48.359719] I [event-epoll.c:629:event_dispatch_epoll_worker] 0-epoll: Started thread with index 2
[2015-08-01 21:40:48.360606] I [rpcsvc.c:2213:rpcsvc_set_outstanding_rpc_limit] 0-rpc-service: Configured rpc.outstanding-rpc-limit with value 64
[2015-08-01 21:40:48.360679] W [options.c:936:xl_opt_validate] 0-nova-server: option 'listen-port' is deprecated, preferred is 'transport.socket.listen-port', continuing with correction
[2015-08-01 21:40:48.361713] E [ctr-helper.c:250:extract_ctr_options] 0-gfdbdatastore: CTR Xlator is disabled.
[2015-08-01 21:40:48.361745] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-pagesize from params.Assigning default value: 4096
[2015-08-01 21:40:48.361762] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-cachesize from params.Assigning default value: 1000
[2015-08-01 21:40:48.361774] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-journalmode from params.Assigning default value: wal
[2015-08-01 21:40:48.361795] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-wal-autocheckpoint from params.Assigning default value: 1000
[2015-08-01 21:40:48.361812] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-sync from params.Assigning default value: normal
[2015-08-01 21:40:48.361825] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-autovacuum from params.Assigning default value: none
[2015-08-01 21:40:48.362666] I [trash.c:2363:init] 0-nova-trash: no option specified for 'eliminate', using NULL
[2015-08-01 21:40:48.362906] E [posix.c:5894:init] 0-nova-posix: Extended attribute trusted.glusterfs.volume-id is absent
[2015-08-01 21:40:48.362922] E [xlator.c:426:xlator_init] 0-nova-posix: Initialization of volume 'nova-posix' failed, review your volfile again
[2015-08-01 21:40:48.362930] E [graph.c:322:glusterfs_graph_init] 0-nova-posix: initializing translator failed
[2015-08-01 21:40:48.362956] E [graph.c:661:glusterfs_graph_activate] 0-graph: init failed
[2015-08-01 21:40:48.363612] W [glusterfsd.c:1219:cleanup_and_exit] (--> 0-: received signum (0), shutting down

EDIT # 2

OK, então um problema parece estar com o atributo estendido não estando presente no sistema de arquivos do brick montado. Este comando é para corrigir isso:

$ grep volume-id /var/lib/glusterd/vols/nova/info | cut -d= -f2 | sed 's/-//g'
f0d72d64288c4e729c532d16ce5687ac
$ setfattr -n trusted.glusterfs.volume-id -v 0xf0d72d64288c4e729c532d16ce5687ac /data/glusterfs/sde1

No entanto, ainda estou recebendo o aviso acima sobre o atributo estar ausente:

[2015-08-01 18:44:50.481350] E [posix.c:5894:init] 0-nova-posix: Extended attribute trusted.glusterfs.volume-id is absent

Saída completa do glusterd restart :

[2015-08-01 22:03:41.467668] I [MSGID: 100030] [glusterfsd.c:2294:main] 0-/usr/sbin/glusterfsd: Started running /usr/sbin/glusterfsd version 3.7.0 (args: /usr/sbin/glusterfsd -s icicle11 --volfile-id nova.icicle11.data-glusterfs-sde1-brick -p /var/lib/glusterd/vols/nova/run/icicle11-data-glusterfs-sde1-brick.pid -S /var/run/gluster/d0a51f364706915faa35c6cca46e9ce6.socket --brick-name /data/glusterfs/sde1/brick -l /var/log/glusterfs/bricks/data-glusterfs-sde1-brick.log --xlator-option *-posix.glusterd-uuid=5e09f3ec-bfbc-490b-bd93-8e083e8ebd05 --brick-port 49155 --xlator-option nova-server.listen-port=49155)
[2015-08-01 22:03:41.514878] I [event-epoll.c:629:event_dispatch_epoll_worker] 0-epoll: Started thread with index 1
[2015-08-01 22:04:00.334285] I [graph.c:269:gf_add_cmdline_options] 0-nova-server: adding option 'listen-port' for volume 'nova-server' with value '49155'
[2015-08-01 22:04:00.334330] I [graph.c:269:gf_add_cmdline_options] 0-nova-posix: adding option 'glusterd-uuid' for volume 'nova-posix' with value '5e09f3ec-bfbc-490b-bd93-8e083e8ebd05'
[2015-08-01 22:04:00.334518] I [server.c:392:_check_for_auth_option] 0-/data/glusterfs/sde1/brick: skip format check for non-addr auth option auth.login./data/glusterfs/sde1/brick.allow
[2015-08-01 22:04:00.334529] I [server.c:392:_check_for_auth_option] 0-/data/glusterfs/sde1/brick: skip format check for non-addr auth option auth.login.a9c47852-7dcf-4f89-80e5-110101943f36.password
[2015-08-01 22:04:00.334540] I [event-epoll.c:629:event_dispatch_epoll_worker] 0-epoll: Started thread with index 2
[2015-08-01 22:04:00.335316] I [rpcsvc.c:2213:rpcsvc_set_outstanding_rpc_limit] 0-rpc-service: Configured rpc.outstanding-rpc-limit with value 64
[2015-08-01 22:04:00.335371] W [options.c:936:xl_opt_validate] 0-nova-server: option 'listen-port' is deprecated, preferred is 'transport.socket.listen-port', continuing with correction
[2015-08-01 22:04:00.336170] E [ctr-helper.c:250:extract_ctr_options] 0-gfdbdatastore: CTR Xlator is disabled.
[2015-08-01 22:04:00.336190] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-pagesize from params.Assigning default value: 4096
[2015-08-01 22:04:00.336197] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-cachesize from params.Assigning default value: 1000
[2015-08-01 22:04:00.336211] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-journalmode from params.Assigning default value: wal
[2015-08-01 22:04:00.336217] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-wal-autocheckpoint from params.Assigning default value: 1000
[2015-08-01 22:04:00.336235] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-sync from params.Assigning default value: normal
[2015-08-01 22:04:00.336241] W [gfdb_sqlite3.h:238:gfdb_set_sql_params] 0-nova-changetimerecorder: Failed to retrieve sql-db-autovacuum from params.Assigning default value: none
[2015-08-01 22:04:00.336951] I [trash.c:2363:init] 0-nova-trash: no option specified for 'eliminate', using NULL
[2015-08-01 22:04:00.337131] E [posix.c:5894:init] 0-nova-posix: Extended attribute trusted.glusterfs.volume-id is absent
[2015-08-01 22:04:00.337142] E [xlator.c:426:xlator_init] 0-nova-posix: Initialization of volume 'nova-posix' failed, review your volfile again
[2015-08-01 22:04:00.337148] E [graph.c:322:glusterfs_graph_init] 0-nova-posix: initializing translator failed
[2015-08-01 22:04:00.337154] E [graph.c:661:glusterfs_graph_activate] 0-graph: init failed
[2015-08-01 22:04:00.337629] W [glusterfsd.c:1219:cleanup_and_exit] (--> 0-: received signum (0), shutting down

    

Answer 1

OK, então parece que eu tive que fazer o seguinte.

1. Adicione o atributo estendido trusted.glusterfs.volume-id - observe que ele precisa estar no diretório /brick , tentei subir de nível e não funcionou

   $ setfattr -n trusted.glusterfs.volume-id -v 0xf0d72d64288c4e729c532d16ce5687ac /data/glusterfs/sde1/brick
   

   NOTA: esse valor para o volume-id vem deste comando:

   $ grep volume-id /var/lib/glusterd/vols/nova/info | cut -d= -f2 | sed 's/-//g'
   f0d72d64288c4e729c532d16ce5687ac
   

2. Reinicie o glusterd

   $ service restart glusterd.service
   

   Se eu observar o log do bloco: /var/log/glusterfs/bricks/data-glusterfs-sde1-brick.log , você verá as mensagens do efeito:

   [2015-08-01 22:28:01.510200] I [login.c:81:gf_auth] 0-auth/login: allowed user names: a9c47852-7dcf-4f89-80e5-110101943f36
   [2015-08-01 22:28:01.510254] I [server-handshake.c:585:server_setvolume] 0-nova-server: accepted client from icicle07.td.teradata.com-44127-2015/08/01-21:08:06:639278-nova-client-19-0-0 (version: 3.7.0)
   [2015-08-01 22:28:01.510584] I [login.c:81:gf_auth] 0-auth/login: allowed user names: a9c47852-7dcf-4f89-80e5-110101943f36
   [2015-08-01 22:28:01.510614] I [server-handshake.c:585:server_setvolume] 0-nova-server: accepted client from icicle08.td.teradata.com-7291-2015/07/02-00:22:13:514999-nova-client-19-0-0 (version: 3.7.0)
   [2015-08-01 22:28:01.513443] I [login.c:81:gf_auth] 0-auth/login: allowed user names: a9c47852-7dcf-4f89-80e5-110101943f36
   

   Agora, enquanto vejo o bloco que vejo, ele está sendo sincronizado com o restante do cluster:

   $ while [ 1 ]; do du -sh /data/glusterfs/sde1/brick; sleep 30; done
   38G /data/glusterfs/sde1/brick
   40G /data/glusterfs/sde1/brick
   41G /data/glusterfs/sde1/brick 
   

3. Quando isso estiver concluído, execute um comando de recuperação para verificar novamente as coisas.

   $ gluster volume heal nova full
   

Detalhes adicionais

Também vi essas mensagens logo depois que reiniciei o glusterd :

[2015-08-01 22:27:56.882271] W [graph.c:357:_log_if_unknown_option] 0-nova-quota: option 'timeout' is not recognized
[2015-08-01 22:27:56.882303] W [graph.c:357:_log_if_unknown_option] 0-nova-trash: option 'brick-path' is not recognized
Final graph:
+------------------------------------------------------------------------------+
  1: volume nova-posix
  2:     type storage/posix
  3:     option glusterd-uuid 5e09f3ec-bfbc-490b-bd93-8e083e8ebd05
  4:     option directory /data/glusterfs/sde1/brick
  5:     option volume-id f0d72d64-288c-4e72-9c53-2d16ce5687ac
  6: end-volume
  7:
  8: volume nova-trash
  9:     type features/trash
 10:     option trash-dir .trashcan
 11:     option brick-path /data/glusterfs/sde1/brick
 12:     option trash-internal-op off
 13:     subvolumes nova-posix
 14: end-volume
 15:
 16: volume nova-changetimerecorder
 17:     type features/changetimerecorder
 18:     option db-type sqlite3
 19:     option hot-brick off
 20:     option db-name brick.db
 21:     option db-path /data/glusterfs/sde1/brick/.glusterfs/
 22:     option record-exit off
 23:     option ctr_link_consistency off
 24:     option record-entry on
 25:     option ctr-enabled off
 26:     option record-counters off
 27:     subvolumes nova-trash
 28: end-volume
 29:
 30: volume nova-changelog
 31:     type features/changelog
 32:     option changelog-brick /data/glusterfs/sde1/brick
 33:     option changelog-dir /data/glusterfs/sde1/brick/.glusterfs/changelogs
 34:     option changelog-barrier-timeout 120
 35:     subvolumes nova-changetimerecorder
 36: end-volume
 37:
 38: volume nova-bitrot-stub
 39:     type features/bitrot-stub
 40:     option export /data/glusterfs/sde1/brick
 41:     subvolumes nova-changelog
 42: end-volume
 43:
 44: volume nova-access-control
 45:     type features/access-control
 46:     subvolumes nova-bitrot-stub
 47: end-volume
 48:
 49: volume nova-locks
 50:     type features/locks
 51:     subvolumes nova-access-control
 52: end-volume
 53:
 54: volume nova-upcall
 55:     type features/upcall
 56:     option cache-invalidation off
 57:     subvolumes nova-locks
 58: end-volume
 59:
 60: volume nova-io-threads
 61:     type performance/io-threads
 62:     subvolumes nova-upcall
 63: end-volume
 64:
 65: volume nova-barrier
 66:     type features/barrier
 67:     option barrier disable
 68:     option barrier-timeout 120
 69:     subvolumes nova-io-threads
 70: end-volume
 71:
 72: volume nova-index
 73:     type features/index
 74:     option index-base /data/glusterfs/sde1/brick/.glusterfs/indices
 75:     subvolumes nova-barrier
 76: end-volume
 77:
 78: volume nova-marker
 79:     type features/marker
 80:     option volume-uuid f0d72d64-288c-4e72-9c53-2d16ce5687ac
 81:     option timestamp-file /var/lib/glusterd/vols/nova/marker.tstamp
 82:     option xtime off
 83:     option gsync-force-xtime off
 84:     option quota off
 85:     option inode-quota off
 86:     subvolumes nova-index
 87: end-volume
 88:
 89: volume nova-quota
 90:     type features/quota
 91:     option volume-uuid nova
 92:     option server-quota off
 93:     option timeout 0
 94:     option deem-statfs off
 95:     subvolumes nova-marker
 96: end-volume
 97:
 98: volume nova-worm
 99:     type features/worm
100:     option worm off
101:     subvolumes nova-quota
102: end-volume
103:
104: volume nova-read-only
105:     type features/read-only
106:     option read-only off
107:     subvolumes nova-worm
108: end-volume
109:
110: volume /data/glusterfs/sde1/brick
111:     type debug/io-stats
112:     option latency-measurement off
113:     option count-fop-hits off
114:     subvolumes nova-read-only
115: end-volume
116:
117: volume nova-server
118:     type protocol/server
119:     option transport.socket.listen-port 49155
120:     option rpc-auth.auth-glusterfs on
121:     option rpc-auth.auth-unix on
122:     option rpc-auth.auth-null on
123:     option transport-type tcp
124:     option auth.login./data/glusterfs/sde1/brick.allow a9c47852-7dcf-4f89-80e5-110101943f36
125:     option auth.login.a9c47852-7dcf-4f89-80e5-110101943f36.password XXXXXXX
126:     option auth.addr./data/glusterfs/sde1/brick.allow *
127:     subvolumes /data/glusterfs/sde1/brick
128: end-volume
129:
+------------------------------------------------------------------------------+

Confirmando os atributos estendidos

Você pode usar o seguinte comando para ver quais atributos estão presentes:

$ getfattr -d -m . -e hex /data/glusterfs/sde1/brick
getfattr: Removing leading '/' from absolute path names
# file: data/glusterfs/sde1/brick
trusted.afr.dirty=0x000000000000000000000000
trusted.afr.nova-client-18=0x000000000000000000000000
trusted.afr.nova-client-19=0x000000000000000000000000
trusted.gfid=0x00000000000000000000000000000001
trusted.glusterfs.dht=0x00000001000000004ccccccb66666663
trusted.glusterfs.dht.commithash=0x3000
trusted.glusterfs.volume-id=0xf0d72d64288c4e729c532d16ce5687ac

Referências

• Gluster 3.4: Restauração de Tijolos - Substitua o Servidor Caiu
• Substituindo um bloco no GlusterFS 3.4.0