We are planning to go for Highly Available DataStage environment.
We have DataStage 8.0 PE running on Sun Sparc T5220 on Unix.
What are the different options available? What are the advantages and disadvantages?
For automatic switching, what H/W and S/W are needed OR DataStage supports?
How Active-Active or Active-Passive will work for us in this scenario? (This is new to me)
Regards,
Ajit
DataStage server Failover for High Availability
Moderators: chulett, rschirm, roy
Making DataStage run correctly in a HA environment is not a simple issue.
Quite a bit depends upon how the users defines High Availability. Being able to restart jobs that were running when the machine died (and failed-over) makes things easier. If that is not the case, i.e. job level is not granular enough, then one needs to design the whole DataStage environment for failover.
Does DataStage use files or datasets - and if so, are they local to the server or on a SAN? If local, are the disks dual ported to allow access when the server has failed (if neither of the two cases above can be done, then HA is unlikely to function).
Implementing a duplicate, hot-standy system with failover to SAN data and HA-Database installations can be done with a fair amoutn of planning. All other options will require quite a bit more investment in planning time and will require quite a bit of maintenance to remain functional. Too many HA systems that I've seen work well in planning and testing but, when the first failover happens after a year in production, tend to fail spectacularly.
Quite a bit depends upon how the users defines High Availability. Being able to restart jobs that were running when the machine died (and failed-over) makes things easier. If that is not the case, i.e. job level is not granular enough, then one needs to design the whole DataStage environment for failover.
Does DataStage use files or datasets - and if so, are they local to the server or on a SAN? If local, are the disks dual ported to allow access when the server has failed (if neither of the two cases above can be done, then HA is unlikely to function).
Implementing a duplicate, hot-standy system with failover to SAN data and HA-Database installations can be done with a fair amoutn of planning. All other options will require quite a bit more investment in planning time and will require quite a bit of maintenance to remain functional. Too many HA systems that I've seen work well in planning and testing but, when the first failover happens after a year in production, tend to fail spectacularly.
</a>