Actually I create logs and writer through Namespace:
Namespace namespace = ...
if (!namespace.logExists(logName)) {
namespace.createLog(logName);
}
DistributedLogManager logManager = namespace.openLog(logName);
AsyncLogWriter asyncLogWriter =
FutureUtils.result(logManager.openAsyncLogWriter());
Then I take a LogReader and start reads
// Actually reader is opened just one time and then reused.
LogReader reader = logManager.openLogReader(DLSN.NonInclusiveLowerBound);
LogRecordWithDLSN record;
while ((record = reader.readNext(!blocking)) != null) {
... business logic ... (i was expecting to get a null and exit from
reading after some timeout)
}
That reader.readNext blocks until I write something into the log:
// Generate the record
ByteBuf buf = Unpooled.wrappedBuffer(os.getByteArray(), 0, os.size());
LogRecord record = new LogRecord(System.currentTimeMillis(), buf);
// Actual write request
final CompletableFuture<DLSN> write = asyncLogWriter.write(record);
Il giorno mar 26 feb 2019 alle ore 12:44 Sijie Guo <guosi...@gmail.com> ha
scritto:
> On Tue, Feb 26, 2019 at 6:37 PM Lothruin Mirwen <lothruin.mir...@gmail.com
> >
> wrote:
>
> > Hi BookKeepers!
> >
> > I have an issue with DL LogReader.readNext(false) [ie: blocking]
> >
> > Such call should "block until return a record if there are records in the
> > stream (aka catching up). Otherwise it would wait up to {@link
> > DistributedLogConfiguration#getReadAheadWaitTime()} milliseconds and
> return
> > null if there isn't any more records in the stream." (from javadoc).
> >
> > It normally works but with empty logs (no data ever written to DL/BK) it
> > blocks forever. It seems that with no data it never "catchUp" and wait
> > forever.
> >
>
> How do you create the empty logs? Did you just open a log writer without
> writing records, or you use Namespace to create a log?
>
> The difference is 1) first approach will have a log segment with zero
> entries; 2) second approach might not even have any log segments.
>
> I guess the difference between 1) and 2) cause the differences at handling
> `readNext(false)`. If you can share more details about your setup, I have a
> better idea of the direction at looking into the problem.
>
>
> > As workaround I forcefully write a dummy record into the log. I'm missing
> > something? Is there any missing configuration I need to provide to BK?
> >
> > Best regards
> >
> > Diego Salvi
> >
>
