一则 gfs2 集群文件系统无法挂载的解决案例

原文地址：http://klwang.info/a-case-of-dlm-controld-and-gfs/

在使用 cman + gfs2 方案时，无法挂载文件系统，日志如下：dlm_controld.log 日志信息

10:22:49 dlm_controld fenced_domain_info error -1

message 日志信息

10:11:08 h4-61 kernel: GFS2: fsid=: Trying to join cluster "lock_dlm", "cluster:fs"
10:11:08 h4-61 kernel: dlm: Using TCP for communications

fenced.log 日志信息

10:25:26 fenced fencing node node3    10:25:26 fenced fence node3 dev 0.0 agent none result: error no method

用strace 查看 dlm_controld 的系统调用，发现一直循环出现下面的信息

socket(PF_FILE, SOCK_STREAM, 0)         = 22
connect(22, {sa_family=AF_FILE, path=@"fenced_query_sock"}, 20) = 0
write(22, "\355\34\341\17\1\0\1\0\6\0\0\0\0\0\0\0 \0\) = 32
read(22, "\355\34\341\17\1\0\1\0\6\0\0\0\0\0\0\0008) = 56
close(22)                               = 0
poll([{fd=5, events=POLLIN}, {fd=8, events=POLLIN}, ) = 0 (Timeout)

查阅了相关资料后，最终才发现是 分布式锁 dlm 的问题，加入下面参数到 cluster.conf中，问题解决

<dlm enable_fencing="0"/>

原来dlm_controld一直在等待fencing，而目前集群中没有配置fencing设备，所以导致了文件系统无法挂载修改后，最终结果如下：message 日志信息

10:23:32 kernel: GFS2: fsid=: Trying to join cluster "lock_dlm", "cluster:fs"
10:23:32 kernel: dlm: Using TCP for communications    10:23:32 kernel: GFS2: fsid=cluster:fs.0: Joined cluster. Now mounting FS...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=0, already locked for use
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=0: Looking at journal...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=0: Done
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=1: Trying to acquire journal lock...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=1: Looking at journal...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=1: Done
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=2: Trying to acquire journal lock...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=2: Looking at journal...
10:23:32 kernel: GFS2: fsid=cluster:fs.0: jid=2: Done

dlm_controld.log 日志信息

10:11:08 dlm_controld uevent: add@/kernel/dlm/fs    10:11:08 dlm_controld kernel: add@ fs    10:11:08 dlm_controld uevent: online@/kernel/dlm/fs    10:11:08 dlm_controld kernel: online@ fs    10:11:08 dlm_controld dlm:ls:fs conf 1 1 0 memb 1 join 1 left    10:11:08 dlm_controld fs add_change cg 1 joined nodeid 1
10:11:08 dlm_controld fs add_change cg 1 we joined    10:11:08 dlm_controld fs add_change cg 1 counts member 1 joined 1 remove 0 failed 0

ok, 问题顺利解决附： 其他有用的 dlm 相关参数protocol
通信协议可以为：tcp, sctp, detect which
协议的选择和 rrp_mode 相关，如果为none，则使用tcp
默认值 detect

<dlm protocol="detect"/>

enable_fencing
encing恢复依赖性开关，默认开启

<dlm enable_fencing="1"/>

enable_quorum
quorum恢复依赖性开关，默认关闭

<dlm enable_quorum="0"/>

enable_deadlk
死锁检测开关，默认关闭

<dlm enable_deadlk="0"/>

参考资料dlm_controld man page