How does SPP select which SPS nodes to use when hosts are unavailable?
If an SPS appliance goes down for any reason, SPP will continue to serve end-users sessions to that now unavailable SPS appliance. This affects disaster recovery purposes as there is no way to remove an SPS appliance from being served to end-users. In the instance of a single-SPS-appliance failure, this leads to end-users being directed to "unavailable" appliances.
The behavior has been changed in the SPP 6.0 release.
1. All SPP and SPS nodes are up and functioning normally a. SPP will query SPS for cluster information and calculate which SPS node is best fit for launching a session
2. All SPP nodes are up and functioning, 1 or more SPS managed hosts are down a. SPP will query SPS for cluster information, detect that 1 or more managed hosts are down and calculate which remaining SPS node is best fit for launching a session. b. Once the offline managed hosts come back online, they will be added back into the available SPS nodes for launching sessions.
3. All SPP nodes are up and functioning, SPS CMS is down and 1 or more managed hosts are still online a. SPP will no longer be able to query SPS for cluster information. Only the CMS is capable of returning information for the entire cluster b. SPP will detect that the CMS is down and based on the last good SPS cluster information it has, will determine the best fit for launching a session. c. In this scenario, if another SPS managed host goes down before the CMS comes back online, SPP may still try to launch a session on the down SPS managed host. The reason for this is because SPP cannot get current SPS cluster information until the CMS comes back online. d. Session launching may or may not work depending on the current state of the other SPS managed hosts. This situation can be temporarily mitigated by defining managed network rules which point the SPP appliances to the available SPS appliances.
4. All SPP nodes are up and functioning, SPS CMS is down and no other managed hosts are available. a. SPP will show the user an error message when they try to launch a session indicating that the sessions module is down. No session can be launch in this scenario. 5. For all variations of SPP nodes including the primary being online or offline, session launching follows the same SPS scenarios above from the remaining SPP nodes.
The only other scenarios that can be added to the above descriptions are introducing a search primary SPS nodes into the SPS cluster. A search primary is not capable of recording sessions and will not be recognized as a session capable appliance. This includes the CMS being configured as a search primary as well. If a CMS is defined as a search primary and there are no other SPS nodes in the cluster, it is the same as if there are no available SPS nodes. This will produce an error when the user tries to launch a session.