Hi Stack,I started to migrate to 0.2.0 anyway in hopes of solving the
problem (and getting the 0.2.0 benefits at the same time).
I can tell you what I was seeing though. With the default lease timeout, it
was always the first region transition. After I upped the lease timeout much
higher (10 minutes), it managed to get through almost all regions (with the
same amount of pause in between). However, the second to last region took
slightly longer than 10 minutes and I got the same effect.
I'll let you know how 0.2.0 is going - I'm having some EC2 related issues
with respect to binding to IPs. More on that later...
Danny
On Wed, Jul 23, 2008 at 3:50 PM, stack <[EMAIL PROTECTED]> wrote:
> I'd expect a little pause as client goes and sets up scanner on the next
> region but your log seems to be showing the client waiting so long that when
> it eventually gets through, server says I know nothing of the lease. What
> if you up the client lease timeout? I wonder if filters are tripping things
> up? Is it the first region transition or sometime after the first/second
> transition? Were 'other scanners on other tasks' that were frozen trying
> to get the same region? Do you know the problematic region? You need to
> pull on the reporter.progress in your map task if its completion takes
> longer than the TaskTracker lease interval.
>
> St.Ack
>
>
> Daniel Leffel wrote:
>
>> Breakthrough in the mystery?
>>
>> The key that it fails on just happens to be the last key in the region....
>>
>> Is opening the scanner on the new region the problem?
>>
>>
>>
>> On Tue, Jul 22, 2008 at 12:18 PM, Daniel Leffel <[EMAIL PROTECTED]
>> >wrote:
>>
>>
>>
>>> OK. So, I didn't notice the IOE subclassing, so that explained the fact I
>>> wasn't catching it.
>>>
>>> However, somethign is still definitely going on. The scanner at some
>>> point
>>> gets "stuck". See the next's being called:
>>> 2008-07-22 13:34:18,274 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: FilterNotNull on
>>> cols:{hierarcy:[EMAIL PROTECTED] Result = false
>>> 2008-07-22 13:34:18,284 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,285 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: FilterNotNull on
>>> cols:{hierarcy:[EMAIL PROTECTED] Result = false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 1121874. Result: false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:1121874. Result = false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 1121874. Result: false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,292 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:1121874, colKey: hierarcy:immediate_parents, data: [EMAIL PROTECTED]
>>> Result = false
>>> 2008-07-22 13:34:18,324 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 11218740. Result: false
>>> 2008-07-22 13:34:18,324 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,325 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:11218740. Result = false
>>> 2008-07-22 13:34:18,325 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 11218740. Result: false
>>> 2008-07-22 13:34:18,325 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,325 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:11218740, colKey: hierarcy:immediate_parents, data: [EMAIL PROTECTED]
>>> Result = false
>>> 2008-07-22 13:34:18,334 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,334 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: FilterNotNull on
>>> cols:{hierarcy:[EMAIL PROTECTED] Result = false
>>> 2008-07-22 13:34:18,744 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Successfully read a
>>> sub-filter of type: org.apache.hadoop.hbase.filter.StopRowFilter
>>> 2008-07-22 13:34:18,745 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Successfully read a
>>> sub-filter of type: org.apache.hadoop.hbase.filter.StopRowFilter
>>> 2008-07-22 13:34:18,745 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Resetting.
>>> 2008-07-22 13:34:18,758 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 11391639. Result: false
>>> 2008-07-22 13:34:18,758 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,758 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:11391639. Result = false
>>> 2008-07-22 13:34:18,759 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 11391639. Result: false
>>> 2008-07-22 13:34:18,759 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,759 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:11391639, colKey: hierarcy:immediate_parents, data: [EMAIL PROTECTED]
>>> Result = false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 1139164. Result: false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:1139164. Result = false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.StopRowFilter:
>>> Filter result for rowKey: 1139164. Result: false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,779 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: Filter on
>>> rowKey:1139164, colKey: hierarcy:immediate_parents, data: [EMAIL PROTECTED]
>>> Result = false
>>> 2008-07-22 13:34:18,782 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter:
>>> this.filterAllRemaining
>>> is now: false
>>> 2008-07-22 13:34:18,783 DEBUG
>>> org.apache.hadoop.hbase.filter.WhileMatchRowFilter: FilterNotNull on
>>> cols:{hierarcy:[EMAIL PROTECTED] Result = false
>>>
>>>
>>> Then at some point, next just hangs and the next entry in the log is
>>> expiration:
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -4780003109149242775/-4780003109149242775
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -4780003109149242775 lease expired
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -8543632393752004827/-8543632393752004827
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -8543632393752004827 lease expired
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -3834469434163616397/-3834469434163616397
>>> 2008-07-22 13:44:18,015 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -3834469434163616397 lease expired
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -6363251876420626143/-6363251876420626143
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -6363251876420626143 lease expired
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -4332402553799574820/-4332402553799574820
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -4332402553799574820 lease expired
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> 7541512102009958038/7541512102009958038
>>> 2008-07-22 13:44:28,024 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> 7541512102009958038 lease expired
>>> 2008-07-22 13:44:28,025 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> -6673102874800967750/-6673102874800967750
>>> 2008-07-22 13:44:28,025 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> -6673102874800967750 lease expired
>>> 2008-07-22 13:44:28,025 INFO org.apache.hadoop.hbase.Leases:
>>> regionserver/0.0.0.0:60020.leaseChecker lease expired
>>> 9143807894343107328/9143807894343107328
>>> 2008-07-22 13:44:28,025 INFO org.apache.hadoop.hbase.HRegionServer:
>>> Scanner
>>> 9143807894343107328 lease expired
>>>
>>> Now interestingly, the map task doesn't know about the expired lease
>>> until
>>> a number of minutes later when it finally tries to call next:
>>> 2008-07-22 14:54:29,062 INFO org.apache.hadoop.ipc.Server: IPC Server
>>> handler 8 on 60020, call next(-4780003109149242775) from
>>> 10.252.242.159:60900: error:
>>> org.apache.hadoop.hbase.UnknownScannerException: Name:
>>> -4780003109149242775
>>> org.apache.hadoop.hbase.UnknownScannerException: Name:
>>> -4780003109149242775
>>> at
>>> org.apache.hadoop.hbase.HRegionServer.next(HRegionServer.java:1506)
>>> at sun.reflect.GeneratedMethodAccessor3.invoke(Unknown Source)
>>> at
>>>
>>> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
>>> at java.lang.reflect.Method.invoke(Method.java:597)
>>> at
>>> org.apache.hadoop.hbase.ipc.HbaseRPC$Server.call(HbaseRPC.java:413)
>>> at org.apache.hadoop.ipc.Server$Handler.run(Server.java:901)
>>> 2008-07-22 14:54:29,070 INFO org.apache.hadoop.ipc.Server: IPC Server
>>> handler 3 on 60020, call close(-4780003109149242775) from
>>> 10.252.242.159:60900: error:
>>> org.apache.hadoop.hbase.UnknownScannerException: -4780003109149242775
>>> org.apache.hadoop.hbase.UnknownScannerException: -4780003109149242775
>>> at
>>> org.apache.hadoop.hbase.HRegionServer.close(HRegionServer.java:1616)
>>> at sun.reflect.GeneratedMethodAccessor11.invoke(Unknown Source)
>>> at
>>>
>>> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
>>> at java.lang.reflect.Method.invoke(Method.java:597)
>>> at
>>> org.apache.hadoop.hbase.ipc.HbaseRPC$Server.call(HbaseRPC.java:413)
>>> at org.apache.hadoop.ipc.Server$Handler.run(Server.java:901)
>>>
>>>
>>> Even more interestingly is that the other scanners for the other tasks
>>> were
>>> also "frozen" during the time and their leases expire too.
>>>
>>> Inside the iteration, it's pretty simple stuff happening - string
>>> manipulation mostly and a put into a bdb - takes <10ms.
>>>
>>> The only other suspect piece of code is a reporter.progress() which I had
>>> to put or else the task tracker thinks the task is dead. I've not had
>>> trouble with reporter.progress() to date...
>>>
>>> Any thoughts?
>>>
>>>
>>> On Tue, Jul 22, 2008 at 11:22 AM, stack <[EMAIL PROTECTED]> wrote:
>>>
>>>
>>>
>>>> Hey D:
>>>>
>>>> TableInputFormat#TableRecordReader#next looks like it lets out
>>>> IOExceptions. UnknownScannerException subclasses IOE so I'd think it
>>>> should
>>>> bubble up into your map task.
>>>>
>>>> Paste in the exception you found in your logs. That might help figure
>>>> whats going on.
>>>>
>>>> St.Ack
>>>>
>>>>
>>>> Daniel Leffel wrote:
>>>>
>>>>
>>>>
>>>>> I've had a little bit of weird behavior.
>>>>>
>>>>> I am opening a scanner in the configure method of a Map task to load a
>>>>> simple little in-memory map (I'd love to this with in-memory column
>>>>> stores,
>>>>> but that's another story ;-).
>>>>>
>>>>> Anyway, the scanner iterates over 4 million rows or so and I've
>>>>> verified
>>>>> next is being called multiple times per second. However, at the same
>>>>> record
>>>>> every time (after around 6 minutes of iteration), the scanner throws an
>>>>> UnknownScannerException in the log (I see it in the log), but rather
>>>>> than
>>>>> throw the exception to the Map task, next just returns false and the
>>>>> loop
>>>>> ends (making it exceptionally difficult to detect if the scanner
>>>>> finished
>>>>> scanning or not). Increasing the regionserver timeout has most
>>>>> definitely
>>>>> solved the issue, but shouldn't calling next reset the timeout anyway?
>>>>>
>>>>> Such a frustrating problem!
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>>
>
>
