Hmm...
Here's what I get when I try to wrap a scanner in my own Scanner implementation:
java.io.IOException: InternalScanner implementation is expected to be
HRegion.RegionScanner.
at
org.apache.hadoop.hbase.regionserver.HRegionServer.next(HRegionServer.java:2023)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:616)
at
org.apache.hadoop.hbase.ipc.WritableRpcEngine$Server.call(WritableRpcEngine.java:314)
at
org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:1225)
Indeed in HRegionServer.next(...) I see this:
...
InternalScanner s = this.scanners.get(scannerName);
...
// Call coprocessor. Get region info from scanner.
HRegion region = null;
if (s instanceof HRegion.RegionScanner) {
HRegion.RegionScanner rs = (HRegion.RegionScanner) s;
region = getRegion(rs.getRegionName().getRegionName());
} else {
throw new IOException("InternalScanner implementation is expected " +
"to be HRegion.RegionScanner.");
}
Forcing all scanners created in {pre|post}OpenScanner to be subclasses of
RegionScanner is not right.
I see why it's done this way, though (after all only the scanner knows its
region).
So I guess either a scanner's region should be stored somewhere else, or there
should be a RegionScanner interface that includes:
public HRegionInfo getRegionName();
Here a simple isolates test for this:
public class ScannerTest extends BaseRegionObserver {
@Override
public InternalScanner postScannerOpen(final
ObserverContext<RegionCoprocessorEnvironment> c,
final Scan scan, final
InternalScanner s) throws IOException
{
// just wrap the passed scanner
return new InternalScanner() {
private InternalScanner del;
{ this.del = s;}
public boolean next(List<KeyValue> results) throws IOException
{
return del.next(results);
}
public boolean next(List<KeyValue> result, int limit) throws
IOException
{
return del.next(result, limit);
}
public void close() throws IOException {
del.close();
}
};
}
}
-- Lars
________________________________
From: Andrew Purtell <[email protected]>
To: "[email protected]" <[email protected]>; lars hofhansl
<[email protected]>
Sent: Tuesday, August 9, 2011 9:50 AM
Subject: Re: Allow RegionCoprocessorEnvironment to register custom scanners?
Great! I'm glad you have everything you need Lars. If at some point you become
stuck because there is indeed some control or API surface missing, please write
back!
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein (via
Tom White)
----- Original Message -----
> From: lars hofhansl <[email protected]>
> To: "[email protected]" <[email protected]>; Andrew Purtell
> <[email protected]>
> Cc:
> Sent: Monday, August 8, 2011 7:53 PM
> Subject: Re: Allow RegionCoprocessorEnvironment to register custom scanners?
>
> I see.I just didn't see how you could communicate any information to the
> server via a Scan, but now I see Scan.setAttribute(...).
>
> Thanks Andy.
>
> -- Lars
>
>
>
> ________________________________
> From: Andrew Purtell <[email protected]>
> To: "[email protected]" <[email protected]>; lars
> hofhansl <[email protected]>
> Sent: Monday, August 8, 2011 5:50 PM
> Subject: Re: Allow RegionCoprocessorEnvironment to register custom scanners?
>
> The RegionObserver already wraps all of the scanner operations.
> RegionObserver.preScannerOpen can create an InternalScanner and return it
> exactly as you propose with "HRegionServer.addScanner(InternalScanner)
> ".
>
> preScannerOpen takes a Scan object.
>
> Only if preScannerOpen does not return an InternalScanner will the
> RegionServer
> look for a "real" InternalScanner.
>
> So I don't see what addScanner would buy you.
>
> Best regards,
>
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein (via
> Tom White)
>
>
>> ________________________________
>> From: lars hofhansl <[email protected]>
>> To: "[email protected]" <[email protected]>
>> Sent: Monday, August 8, 2011 5:38 PM
>> Subject: Allow RegionCoprocessorEnvironment to register custom scanners?
>>
>> Currently coprocessors can't do any streaming operations.
>>
>> I think that would be a necessary feature to perform long running operations
> on the server (like scans) that in turn could produce a lot of data.
>> GroupBy type aggregates come to mind, but there are many more cases.
>>
>>
>> Somewhere I read about some approach for server side cursors (can't find
> that discussion now).
>> I think a simpler approach would be allowing a coprocessor to register new
> InternalScanners that it could implement,
>> and then have some way of accessing the scanner via the normal ClientScanner
> mechanism.
>> Maybe by just exposing long HRegionServer.addScanner(InternalScanner)
> through RegionServerServices.
>> and adding public ResultScanner getScanner(long scannerId) ... on HTable,
> and similar on all other clients (I don't know anything about the client
> beside the HTable Java client).
>>
>>
>> Or something similar (just making this up here).
>>
>>
>> That way all major parts are already in place (Client Scanners are good in
> performing caching, the coprocessor could just wrap "real" internal
> scanners, etc). The problem is just about how to wire up the parts.
>>
>>
>> Thoughts? Are questions like this better asked on the dev list?
>>
>> Thanks.
>>
>> -- Lars
>>
>>
>>
>
