Can a failed transactor's host ip remain in dynamodb?


#1

Hello,

I had this issue several times and am not sure how to fix it.
Sometimes on a failed transactor, it seems that the machines that try to connect to the db are trying to connect to a transactor ip that does not exist.
If I check the aws.properties of my transactor it states: host=10.0.11.186
Yet I see messages in the peers with errors: Error communicating with HOST 10.0.10.236 on PORT 4334
Is it correct that that ip should be the one from my aws.properties host?
How is this possible, since there is only one transactor running, it should be impossible to receive an other IP which are stored in dynamodb if a transactor registers right? So what could be the cause that the ‘active’ transactor registered in dynamodb is actually a dead transactor.

PS: I waited quite some time by now (weekend), in the meantime the peers have restarted several times and I restarted the transactor 2-3 times as well. The ip they try to connect on still remains ‘10.0.10.236’. What could be wrong?

Kind Regards
Brecht


#2

Apparently sometimes the new transactor did not start flawlessly but we did not get any error logs. In that case the the old transactor seems to be still present in the db and it tries to connect to that one (which does not exist anymore). So the problem was not due to datomic!