Does this still happen after a few dozen queries to 10.2.3.38?
Hmm. it depends of downtime and how long it worked before restart. If both times
is rather small then errors/timeouts go away after ten minutes. If at least one
of that times is greater than half an hour then full restoration could take more
than several hours.
It looks like .40 is finding the key and trying to send it back, so
one possibility is that .40's pooled socket to .38 hasn't realized
that .38 was restarted, so the first few messages might get lost
before it reconnects.
Downtime was long enough, all live nodes say by 'nodeprobe cluster' that .38 is
down.
It works well after restarting whole cluster.
--
Teodor Sigaev E-mail: [email protected]
WWW: http://www.sigaev.ru/
