Hi Stack,
I got the log file from the friend and it's too big to fit to
pastebin. I'm going to forward the log to you (actually a link to a
private server) via private email.
Although his colleague thought they met HBase-2482, he read the log
today and he thinks they were not. It seems the runout of the file
handles was preventing the region server to open up the regions.
How can he verify if some regions are still in-transition? scan the
META table?
Even if they didn't meet 2482, I think it's good to roll HBase 0.20.5
so we can protect others to meet this and the another issues.
Thanks,
Tatsuya
On May 16, 2010, at 8:29 AM, Stack <st...@duboce.net> wrote:
2010/5/15 Tatsuya Kawano <tatsuya6...@gmail.com>:
1. Could HBASE-2482 happen on HBase 0.20.4? We think the answer is
yes.
Yes.
2. How can they take the lost regions back?
See bin/add_table.rb. Restart cluster after running it.
3. How to prevent HBASE-2482 on 0.20.4?
Is it okay to apply the v3 patch against 0.20.4 release?
Yes.
But we are about to roll a 0.20.5. I'll include hbase-2482 in it.
I asked him to put the log files to pastebin. They"ll be ready on
Monday.
This would be good. Then we confirm it indeed hbase-2482.
Thanks Tatsuya,
St.Ack
