Understanding timeouts

All EVA ICS calls rely on timeouts. To understand the timeout model, let us review one of the most complicated examples , which may be faced in production.

Call timeouts

Consider we have a deployment payload with lots of items, which must be deployed to a remote node. Even if the registry auto-flush is disabled on the remote node, the call may not fit the default 5-sec timeout.

In this case, the item deployment call is completed, but the client gets timeout and the further deployment process is stopped.

The schema

To understand the problem, let us review the full path of the above call:

timeouts schema

Configuring the timeouts

All the timeouts, present on the above schema, must be increased to allow the long deployment call be processed without timeout errors.

Client timeout

To let a client wait until a long API call is finished, increase its timeout to the corresponding value, which should be slightly higher than the maximum expected operation timeout. In our example, add -T argument for eva-shell (--timeout if calling EVA Cloud Manager CLI directly):

eva -T 60 cloud deploy payload.yml

Local replication pub/sub timeout

To send a call to another node, Replication service is used. Consider it is deployed with the default id eva.repl.default. Modify the service configuration and set the default timeout to a higher value:

eva svc edit eva.repl.default

# ...
timeout:
    default: 30
# ...

This affects the default timeout of all newly added remote nodes.

Individual remote node timeouts

The remote node timeout for RPC calls can be seen with the command:

eva node list -s

and modified with the command:

eva node edit NODENAME

The timeout can be edited for static nodes only. If the node is not a static one, it should be appended first:

eva node append NODENAME

Bus call timeout

The last timeout in the above call is the local BUS/RT call from the remote replication service to the remote node core. To increase the max allowed timeout, repeat the procedure, described in the previous chapter, on the remote node and increase the default timeout for the remote replication service as well (do not confuse with bus timeout, which is used by BUS/RT sockets for low-level packet processing but not for RPC calls):

eva svc edit eva.repl.default

# ...
timeout:
    default: 30
# ...