RE: Data replication 7.20UC1

Dennis Sorry I missed the original posting, but what are your LRU_MIN_DIRTY and = LRU_MAX_DIRTY set to? On a 7.11 DR environment I have seen them set to = as low as 1 and 2 percent respectively. Keeping low values of = LRU_MIN_DIRTY and LRU_MAX_DIRTY will enable more LRU writes so there = will not be so much pending disk writes during checkpoint. The other = thing you could look at is to make the logging mode Unbuffered. This = will flush after every commit. HTH Sujit Pal ---------- From: Juan R. Guzman[SMTP:jguzman@transre.com] Sent: Friday, February 14, 1997 8:50 AM To: informix-list@rmy.emory.edu Subject: Re: Data replication 7.20UC1 Dennis de Kock wrote: >=20 > Hello everyone, >=20 > We started to experiment a little with data replication > using two online 7.20 UC.1 servers. Just standard replication > with a primary and secondary server that are identical. > We are experiencing two problems: > - checkpoints do become very long every once and a while > normally they are approx. 3 to 4 seconds, now we saw occasionally > 50 to 170 seconds. As we need continuous, fast response times this > is far too much. > - the database occasionally becomes extremely slow, though the > machine itself shows no sign of excessive activity. >=20 > In my opinion DR should perhaps cause a small overhead on the > machine, but the engine itself should not be influenced in any > way by log based DR. >=20 > -Has anyone experience with data replication with ODS 7.2x ? > -Did you encounter similar problems ? How did you solve them ? > -Are the other DR products of informix more reliable (enterprise > replication) ? >=20 > I'd appreciate if you'd mail a reply. >=20 > Thanks >=20 > Dennis de Kock Are you sure that the checkpoints are taking longer because of DR? I have DR with 7.11 and the only time (from my experience) DR stops a checkpoint is when the primary is waiting for a response from the secondary. If the two machines are configured differently (as far as online is concerned) try using the same config and see what happens. =20 It might be that the secondary is much slower than the primary. Do you have any entries in the online log file regarding DR when the chkpoints take a long time?