hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From 冯宏华 <fenghong...@xiaomi.com>
Subject 答复: 答复: 答复: egionTooBusyException: Above memstore limit
Date Thu, 27 Feb 2014 01:23:08 GMT
bq. I haven't set the number of the regions my self at the beginning.

If no pre-split, there is only one region for each table from the beginning, so all the writes
go to this single region...very possible to incur SocketTimeoutException(0.94) or RegionTooBusyException(0.96+).


You can pre-split when creating the table to make each regionserver to serve at least 10+
regions, that would help a lot in term of write-load balance
________________________________________
发件人: shapoor [esmaili_607@yahoo.com]
发送时间: 2014年2月27日 1:50
收件人: user@hbase.apache.org
主题: Re: 答复: 答复: egionTooBusyException: Above memstore limit

I haven't set the number of the regions my self at the beginning. In 0.94
with region size of 10 gig, I start with one region and after around 250 gig
of saves I see 60 regions are running and somewhere around here the timeout
exception flies around.

java.util.concurrent.ExecutionException: java.net.SocketTimeoutException:
Call to kcs-testhadoop01/192.168.111.210:60020 failed on socket timeout
exception: java.net.SocketTimeou
tException: 60000 millis timeout while waiting for channel to be ready for
read. ch : java.nio.channels.SocketChannel[connected
local=/192.168.111.210:35248 remote=kcs-testhadoop01/192.168.111.210:60020]
...

But hbase continues the jobs and I have now reached over 300 gig of saves.
for each 10000 saves, there are 100 loads in the process. I will do the same
again with 0.96 and let you know but for 0.94 it is still running with only
one exception as I described. but I am sure more will come.

regards.



--
View this message in context: http://apache-hbase.679495.n3.nabble.com/RegionTooBusyException-Above-memstore-limit-tp4056339p4056413.html
Sent from the HBase User mailing list archive at Nabble.com.
Mime
View raw message