understand 11G clusterware logfile(保留策略及修改方式)

在分析一个问题时发现mDNS的LOG所有文件才只保留了几分钟的日志, 我们都知道CRS里的日志多数都是CRS自身滚动保存覆盖的,
我被问到能不能把mDNS日志多保留一段时间? 特意总结了一下CRS日志的保存规则.
CRS的日志是自动管理的,在12c的版本以前日志都在$GI_HOME/log/<node_name>,从12C的版本开始CRS的日志也转移到了ADR中,如同11G时的bdump等日志, 12c以后的log存在<ORACLE_BASE>/crsdata/<node>/output
和<ORACLE_BASE>/diag/crs/<node>/crs 目录中.这里我们内容还是关于11G版本(当前11.2.0.3)
1, 不是自动管理,需要DBA维护

<GRID_HOME>/log/$HOST (has alert$HOST.log)
<GRID_HOME>/log/$HOST/client
<GRID_HOME>/log/diag
<GRID_HOME>/log/diag/asm/user_root/host_*
<GRID_HOME>/log/diag/asmcmd/user_grid/$HOST
<GRID_HOME>/log/diag/asmcmd/user_oracle/$HOST
<GRID_HOME>/log/diag/asmcmd/user_root/$HOST
<GRID_HOME>/log/diag/asmtool/user_root/host_*
<GRID_HOME>/log/diag/clients/user_root/host_*
<GRID_HOME>/log/diag/tnslsnr/$HOST/listener_scan1
<GRID_HOME>/log/diag/tnslsnr/$HOST/listener_scan2
<GRID_HOME>/log/$HOST/racg
<GRID_HOME>/log/$HOST/racg/racgeut
<GRID_HOME>/log/$HOST/racg/racgevtf
<GRID_HOME>/log/$HOST/racg/racgmain
<GRID_HOME>/log/$HOST/srvm
<GRID_HOME>/rdbms/audit

2, 自动管理,当前日志和10份50M的历史日志归档,名称为xxxx.l(01-10),FIFO的方式保存,归档时文件名编号+1,被下一个替换.

<GRID_HOME>/log/$HOST/cssd:

3, 自动管理, 当前日志和10份10M的历史日志归档,名称为xxx.l(01-10),FIFO的方式保存,归档时文件名编号+1,被下一个替换.

<GRID_HOME>/log/$HOST/acfsrepl
<GRID_HOME>/log/$HOST/acfsreplroot
<GRID_HOME>/log/$HOST/acfssec
<GRID_HOME>/log/$HOST/admin
<GRID_HOME>/log/$HOST/agent
<GRID_HOME>/log/$HOST/agent/crsd/ora_oc4j_type_oracle
<GRID_HOME>/log/$HOST/agent/crsd/oraagent_grid
<GRID_HOME>/log/$HOST/agent/crsd/oraagent_oracle
<GRID_HOME>/log/$HOST/agent/crsd/orarootagent_root
<GRID_HOME>/log/$HOST/agent/crsd/scriptagent_grid
<GRID_HOME>/log/$HOST/agent/ohasd
<GRID_HOME>/log/$HOST/agent/ohasd/oraagent_oracle
<GRID_HOME>/log/$HOST/agent/ohasd/oraagent_grid
<GRID_HOME>/log/$HOST/agent/ohasd/oracssdagent_root
<GRID_HOME>/log/$HOST/agent/ohasd/oracssdmonitor_root
<GRID_HOME>/log/$HOST/agent/ohasd/orarootagent_root
<GRID_HOME>/log/$HOST/agent/crsd
<GRID_HOME>/log/$HOST/crflogd
<GRID_HOME>/log/$HOST/crfmond
<GRID_HOME>/log/$HOST/crsd
<GRID_HOME>/log/$HOST/ctssd
<GRID_HOME>/log/$HOST/cvu
<GRID_HOME>/log/$HOST/cvu/cvulog
<GRID_HOME>/log/$HOST/cvu/cvutrc
<GRID_HOME>/log/$HOST/diskmon
<GRID_HOME>/log/$HOST/evmd
<GRID_HOME>/log/$HOST/gipcd
<GRID_HOME>/log/$HOST/gpnpd
<GRID_HOME>/log/$HOST/gnsd
<GRID_HOME>/log/$HOST/mdnsd
<GRID_HOME>/log/$HOST/ohasd

当然滚动的日志大小不一定完全一样取决于最后一次写buffer size, 规则如下:

declare
current size of the file = x
buffer size to be written into the file in the current write call = y
Maximum file size = z
if (x+y > z) then
rotate file
else
write to current file
end if;

回到上面的问题, 能修改默认日志最大大小么?
在11g我发现只有CSSD log可以修改, 如从默认的50M修改成150M, 另外注意到Exadata环境中是建议加大的.

[anbob@root]#  crsctl get css logfilesize
CRS-4676: Successful get logfilesize 52428800 for Cluster Synchronization Services.
[anbob@root]#  crsctl set css logfilesize 152428800
CRS-4686: Successful set logfilesize 152428800 for Cluster Synchronization Services.
[anbob@root]#  crsctl get css logfilesize
CRS-4676: Successful get logfilesize 152428800 for Cluster Synchronization Services.

在12c我发现增加了一个新的选项tracefileopts, 可以自定义其它日志大小.

Usage:
crsctl set tracefileopts {mdns|gpnp|css|crf|crs|ctss|evm|gipc} -filesize  <file_size>[K|k|M|m|G|g] -numsegments <num_of_segments>
and
crsctl get tracefileopts {mdns|gpnp|css|crf|crs|ctss|evm|gipc}

另外附之前写的清理日志的脚本:
Shell: To delete listener log file contains text and xml format
script: purge ADR file