TAFC Utility

This section describes about the TAFC utility checks.

The TAFC utility checks that all of a cluster’s servers are configured such that the cluster starts successfully. At present, the cluster administrator must manually invoke TAFC on all servers. Although all instances do not need to be started at exactly the same moment, they must be allowed to run in parallel. If all instances operate for a reasonable time (several minutes), without displaying failure or returning to the command line, then the cluster is correctly configured. TAFC can then be stopped and start the server instances.

Initially TAFC runs all the TICs for a given context. Some TICs check behavior of the local system. Some use confchk to compare local parameter with equivalent parameters on remote TAFC instances. Since several calls to confchk are made, this may take a few seconds.

If all TICs pass, then TAFC starts a reporter service. This receives and replies to parameter requests from remote confchk calls. Periodically TAFC checks and reruns any TICs that have a repeat frequency.

If at any point a TIC fails, then instability in the cluster exists and TAFC takes its reporter service off line and exits.

Operation

The TAFC mechanism is not 100% deterministic. A slow, but misconfigured, server could be missed in the initial check. Also, if misconfigured server started first, it can freeze out good servers on the same context. With a TAFC instance running on each server, any misconfiguration should become apparent within a few cycles of running the TICs.

This flags up the problem to the cluster administrator who can then diagnose and resolve the issue. TAFC should rapidly find most misconfiguration issues most of the time.

UUIDs

The local reporter service also receives the confchk broadcast. If this were to respond, to its own request, then any check for unequal parameters would always fail.  To prevent this, each instance of TAFC generated a unique, time stamp based, UUID and writes it into the TAFC_UUID_ENV_VAR environment variable. TAFC then can identify and ignore its own broadcasts.

If preferred, a UUID can manually be generated for each server and hard coded into the environment variable. For example,

set TAFC_UUID_ENV_VAR=7bc17c6a-520d-11df-8002-0016175d8499

If this environment variable is not set, then each confchk instance generates its own UUID. That is, all TAFC instances will reply to it.

TAFC Monitor Events

If monitoring is enabled, then TAFC also sends TIC pass or fail events to the monitor. The event category is TAFC and failures are classified as errors. The monitor can thus collate information from several TAFC instances running on different servers. For more information about monitor configuring, see the monitor user guide.

TAFC Services

In windows platform, the TAFC utility can be run as a service in background. There can be independent services started based on different TAFC context. These TAFC services support basic operations like install, start, stop and remove.

By default, TAFC service will be running with default context. You have to pass -c context option to start tafc service for a specific TAFC context. You should be able to run parallel TAFC services with a limitation that only one service for each context.