My high-level scenario is that I have a pool of slave machines that I want
to run tests on, but I need to use two machines for each run [one acting as
a client, one acting as a server]. I would like to have it choose two
machines from the pool and then run the tests on them, keeping other test
runs from using those two machines while they are in use for the first run.
I have set up a main build flow job that kicks off a server build flow job,
which it captures the results of in order to know the machine name of the
server for the run. It also kicks off a client job in parallel that takes
the server machine name as a parameter in order to interact with it.
I use the Throttle Concurrent Builds option to restrict the main build flow
and the server build flow jobs to only running 1 concurrent build per node
[they are in the same category]. This prevents these jobs from running on a
node that already has one of them running on it.
The main build flow job looks essentially like this now:
def serverResult
def clientResult
def serverBuildFlow = {
serverResult = build(
'ServerBuildFlow',
BUILD_CHANGELIST: params['BUILD_CHANGELIST'],
TEST_SUITE: params['TEST_SUITE'],
BRANCH: params['BRANCH']
)
}
def clientBuildFlow = {
clientResult = build(
'Client',
BUILD_CHANGELIST: params['BUILD_CHANGELIST'],
RUN_ON_NODE: build.properties['builtOnStr'],
TEST_SUITE: params['TEST_SUITE'],
SERVER_NODE:
serverResult.build.properties['builtOnStr'],
BRANCH: params['BRANCH']
)
}
guard {
parallel (
serverBuildFlow,
clientBuildFlow
)
} rescue {
build(
'Cleanup',
RUN_ON_NODE: build.properties['builtOnStr'],
BUILD_CHANGELIST: params['BUILD_CHANGELIST'],
TEST_SUITE: params['TEST_SUITE'],
SERVER_NODE:
serverResult.build.properties['builtOnStr'],
BRANCH: params['BRANCH']
)
}
It has two shortcomings:
1. The ServerBuildFlow job starts up the server components and then the
job finishes. This then allows another job to be scheduled on the server
node, which is completely disruptive. I need the server build flow job to
continue running until the client is done with the server, then I need it
to end, so that the machine is freed up to be a client or server again.
2. Because the server build flow job and the client job are started in
parallel, the client job isn't picking up
serverResult.build.properties['builtOnStr'], which it needs to know in
order to connect to the appropriate server machine. Because the server
build flow job needs to be running for the entire duration, and otherwise
the client job doesn't need to wait for the server build flow job, I have
them running in parallel.
Any ideas on how to address these two issues, or just accomplish my
high-level goal in another fashion?
Thanks,
Jeremy
--
You received this message because you are subscribed to the Google Groups
Jenkins Users group.
To unsubscribe from this group and stop receiving emails from it, send an email
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
