I tried the same scenario with standard and Asynchbase client and they're both able to recover from the situation (with default configuration). Interesting that the Asynchbase client recovers A LOT faster than the standard client.
So it looks like a bug to me - unless there's a trick or configuration that I don't know about :-) On Tue, Feb 11, 2014 at 9:15 PM, Kristoffer Sjögren <[email protected]>wrote: > Sure thing. https://issues.apache.org/jira/browse/PHOENIX-41. > > In the meantime ill look through the configuration and try a raw > HConnection and see if the same error occur. > > > On Tue, Feb 11, 2014 at 5:48 PM, James Taylor <[email protected]>wrote: > >> Hi Kristoffer, >> That sounds like a bug to me. One thing to check is your HBase time-out >> parameters (I'd list them, but I don't know what they all are). If you dial >> these down (as HBase tries very hard, again and again to reconnect), is the >> behavior any different. I know at Salesforce we don't have this issue, >> though. Perhaps Lars, Eli, or Simon know what we've configured to have this >> not happen. >> >> Would you mind filing a JIRA on this? >> >> Thanks, >> James >> >> >> On Tue, Feb 11, 2014 at 8:20 AM, Kristoffer Sjögren <[email protected]>wrote: >> >>> Hi >>> >>> If HBase restart while sqlline is connected the client hangs and seem >>> beyond recovery. It doesn't respond to system calls and doesn't . >>> >>> It is not a big deal that sqlline dies, but the same thing happens in >>> the application when using JDBC to communicate with phoenix/hbase. >>> >>> I should mention that my test run a single HBase master on localhost. >>> >>> The following files show logs just after restart when the application >>> tries to reconnect. >>> >>> Application: http://pastebin.com/mWAXyLvw >>> HBase: http://pastebin.com/SWPc7kw8 >>> >>> Is this a bug, or? >>> >>> Cheers, >>> -Kristoffer >>> >>> >>> >>> >>> >>> >>> >> >
