We want to have Active Active DR where both the Spectrum, e-Health DR servers should poll devices/elements.
We have a separate trap receiver server.
Please suggest what is the best way of achieving this.
Currently How we are doing is discovering devices in bot environments which is kind of lame.
Also, we have CA Arcserve so replication of any files etc is not an issue.
Only requirement is both should be active, poll , receive traps etc.traps they are receiving today as we have set it to send to both servers
Just want to get around manual discovery part OR doing in a better way of doing ACTIVE ACTIVE DR Spectrum-eHealth both.
Spectrum and eHealth are on separate servers
For Spectrum you can use the default Fault Tolerant installation. Just schedule back-ups every few hours and all changes are sent automatically from the primary server to the fault tolerant server. You'll still have to sync configuration files for OneClick server.
What you are describing here is not DR, is double installation. DR assumes the secondary sites has the same data and is able to take over when the primary site is down. Basically the same data is sent from primary server to disaster recovery server. What you have is double everyting, thus the double-discovery of devices. You don't even share the same data, as one server could read a value for a monitored parameter and the second one would have completely different timestamp and value, as they are not in sync.
For eHealth, the DR installation is fully described in the documentation. The entire process also. It's not active-active.
Yes, You are right it will be Active Active for both Spectrum and e-Health Server.
Is there a way to achieve this simpler, like if we do on one spectrum server it replicates at the same time/delayed on other server?
Same for e-Health?
Traps is the easy part as we set trap exploder separately which is doing this job fine by sending to 2 destinations.
Main issue is maintaining 2 setups on daily basis is kind of difficult so thinking if there is a better way of doing it?
I have in the past written a script which tests connectivity before adding a device to the device you want to add in both. Then once it confirms the connectivity is there - it adds it to both via REST in Spectrum and nhCommand in eHealth. You can check the result from either to see if it was discovered correctly. This way you maintain one list of devices. We even had a field for 'Spectrum only, or both eHealth and Spectrum' with extra ehealth info such as what interfaces to enable reports for.
I can't think of an out-of-the-box solution.
For traps you can just configure both Spectrum and both eHealth servers to receive the traps as they won't process them unless they are active. The polling is the tricky part. Spectrum can be configured to active poll or just poll from the active landscape.
eHealth is the one with the issue and sadly the 'noisier' one because it polls a lot of SNMP oid's. Not a problem if you don't mind polling everything 2x from eHealth.
Is it possible if you can share script, it looks interesting and points towards solution.
Polling we are OK with Active- Active polling by both set of servers e-Health and Spectrum.
I am also researching more into it.
Actually, we want close to zero downtime that's why this model is suggested and trying to built up same.
High Availability model I think uses cluster setup, Since spectrum/e-health is already installed.
Cluster model will be difficult.
HA for eHealth works only with SunCluster or Veritas Cluster. You have to have a shared disc and SPARC processors. It's active-passive scenario, with only one active.
0 downtime for Spectrum is quite easy. The FT design on the application handles that very well. The tricky part comes when dealing with Report Manager. That one doesn't have/support any type of replication. You have to set it your own with 3rd party. It will be unsupported by CA. I have set a scenario like this using Mysql replication and CABI Federation. It works but it's hard to maintain. To ensure that data is replicated over to the failover node, I set the reports to be scheduled on the stand-by node. Once replication is faulty for some reason, you'll see it the data reports missing.
I don't recommend it but it's possible.