Hi, I have 24 ocfs2 volumes on FC storage on 2 ocfs2 node connected 10gig fiber ethernet as back to back. When I start copy process from the some ocfs2 volumes to another ocfs2 volumes via rsync program, one node crashes (I tried many times). There is no any log message in the log files (kern.log, messages etc.). When I add the "sleep 6" line between each copy process, no problem. But if the sleep parameter is less than 5 again it crashes. I think the reason is default "commit=5" parameter. Maybe DLM gives buffer overflow. I don't know :( Any suggestion? Thanks, -- M.At?f CEYLAN Yurdum Yaz?l?m
"M.Atıf CEYLAN"
2013-Dec-11 09:34 UTC
[Ocfs-users] rebooting after many read/write[SOLVED]
On 09-12-2013 11:22, "M.At?f CEYLAN" wrote:> Hi, > I have 24 ocfs2 volumes on FC storage on 2 ocfs2 node connected 10gig > fiber ethernet as back to back. When I start copy process from the > some ocfs2 volumes to another ocfs2 volumes via rsync program, one > node crashes (I tried many times). There is no any log message in the > log files (kern.log, messages etc.). When I add the "sleep 6" line > between each copy process, no problem. But if the sleep parameter is > less than 5 again it crashes. I think the reason is default "commit=5" > parameter. Maybe DLM gives buffer overflow. I don't know :( > > Any suggestion? > Thanks,Hi all, I solved this issue. I have added "commit=1" mount option to fstab and solved the rebooting problem. Perhaps this is related with dlm buffer overflow. Because there are many volumes on the nodes and read/write operations on them. I think it can't buffered concurrent many transactions. -- M.At?f CEYLAN Yurdum Yaz?l?m
Possibly Parallel Threads
- ocfs2_check_dir_for_entry:2119 ERROR: status = -17
- "layout is NULL", "Failed to get node-uuid for [...] and other errors during rebalancing in 3.3.1
- CentOS 5.0 and php-yaz
- Rebuilding PHP: how do I manage updates?
- Changing the position of label title and rotating the title