Hello,

I have implemented a master / slave configuration for our application and the lucene index.

Currently our config has the master index replicating the index every 1min.

Sometimes when I start the application, I receive a '' error. On startup this is fatal and the application needs to be restarted. I have also noticed that this can happen during the application running, but will only log an error and continue on.

My guess is that when the application starts up, if replication is in process, the current marker may be changing and the application can't determine the correct lucene directory.

My question is if there are any settings that can force my application to retry determining the current lucene directory. I know that if we made the replication time longer that there is a better chance of this not occurring but, I'd like to keep the time quite low for now.

Any help on this issue would be appreciated.

Hi,

there is indeed a known 'hole' in the startup of the FSMasterDirectoryProvider where temporarily no marker file exists. Are you starting master and slave at the same time? How big are your indexes. I would recommend that you make sure that the master is started before the slaves.

As you say, this could also happen during runtime, however, for the slaves it just means that they skip a index synchronization. I guess we should log a warning instead of an error.

Currently there is no settings to change this behavior. As said, starting the master first should help and of course increasing the sync interval.

I created a Jira issue HSEARCH-532.

--Hardy