[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference smurf::ase

Title:ase
Moderator:SMURF::GROSSO
Created:Thu Jul 29 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2114
Total number of notes:7347

2048.0. "OPS and drd failover time" by BRSADV::BUGGENHOUT () Wed May 07 1997 11:37

Hi

I am not a db specialist nor and TCR guru nor a developer so I have the next
questions :

- we have a small OPS db on a drd service and from a NT client application
  we make a connection on 1 of the instances, if this system goes down
   the application returns an error and does a reconnect to the second
   instance. This works fine. The reconnect goes immediate as the   
   listener is accepting the connection but the next query hangs until the
   drd service is failed over. This is logic as during the switch the
   disks in the service are locked so we need to wait until the disks are
   available and so is the db. This takes about 20 to 30 sec, can we make
   this go faster ?

- failover goes fast when we are connected on the failing system. If we
  do a new connect to the instance which is not there the failover to
  the other instance works too but takes about 1,5 min. We have to deal 
  TCP/IP timeouts ??  At the end we are trying to
  connect to a not-running host. Any suggestions how we could make this time
   shorter (what kind of test could be implemented in the application to
   switch faster when host is not there at all ? )

Thanks for some input

regina
T.RTitleUserPersonal
Name
DateLines
2048.1some extra remarksBRSADV::BUGGENHOUTWed May 07 1997 13:4314
we play around and notice the following :

- as said in .0 we have a 'long' failover time for the drd service but even
  longer when we just plug out the cable of the KZPSA.
  In the deamon logfile we see messages as 
  'unable to reserve <diskname>'

  ... then failing over the service anyway but it takes about 3,5 min

- my colleque says she tried the same things on TCR V1.0 and this worked
  a lot faster.
  Are there any parameters we can set to shorten this time-outs ?

Regina