Tim Sell wrote:
...
I understand until recently the logs were filled with warnings about
this during every split while a region was temporarily offline so to
speak? it would figure itself out once the region was opened, but
logging these warnings was removed because it was confusing people?
No. They are still there. They are logged at DEBUG level unless we
exhaust retries. Then you'll get a NSRE out on your client as an error
(That said, they are showing in your logs here because they are logged
at INFO level by hadoop ipc).
This isn't one of those warnings, it's the only
NotServingRegionException in my logs.
I'm using this morning's trunk.
I've attached my master and region server logs, if anyone can help. It
happens around the 2008-08-05 12:06 mark.
I added some blank lines to help you find where the activity stopped.
(there is about a 30 minute gap in the region server logs afterwards
as there were no more input to be compacted, split, etc).
I took a look. Seems fine.
As Andrew asks, the thrift client log would be handy....
But this in your second message is disturbing:
08/08/05 13:56:38 DEBUG client.HConnectionManager$TableServers:
reloading table servers because: HRegionInfo was null or empty in
.META.
We're stuck trying to get .META. info but some rows are damaged; have
null content.
For sure you are at TRUNK from this morning? I thought we printed out
all that was non-empty from the row to give us a clue which row is busted.
What happens if you try and scan using the shell?
> scan '.META.'
St.Ack
[Towards the end there is some irrelevant stuff logged because I was
fiddling in the hbase shell. ignore that]
~Tim.
2008/8/5 Andrew Purtell <[EMAIL PROTECTED]>:
Hello Nickolai,
When your regions unexpectedly closed, is/was there anything
logged to the regionserver and master logs at those times?
As for your problem right now, please consider stopping, moving
the old log files out of the way, and then restarting. Please
then try to access your table. Then, please send the resulting
master and region server logs. I'll be happy to take a look.
Thanks,
- Andy
From: Nickolai Toupikov <[EMAIL PROTECTED]>
Subject: Issue with NotServingRegionException
To: [email protected]
Date: Tuesday, August 5, 2008, 5:47 AM
Hello!
It seems I have a serious problem with HBase. We've
been using if for some time now, had some minor problems with
regions unexpectedly closing once every couple of days, but
this would usually sort itself out after a couple of stops and
starts.
>
