Skip to content

[Slurm] Add retry_delays when waiting for workers #176

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 8, 2021
Merged

[Slurm] Add retry_delays when waiting for workers #176

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3adb656
[Slurm] Add retry_delays when waiting for workers
David96 Sep 30, 2021
27c7053
[Slurm] add default constructor for SlurmManager
David96 Oct 6, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 31 additions & 3 deletions src/slurm.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,11 @@ import Logging.@warn

struct SlurmManager <: ClusterManager
np::Integer
retry_delays
end

struct SlurmException <: Exception
msg
end

function launch(manager::SlurmManager, params::Dict, instances_arr::Array,
Expand Down Expand Up @@ -54,12 +59,13 @@ function launch(manager::SlurmManager, params::Dict, instances_arr::Array,
srun_cmd = `srun -J $jobname -n $np -o "$(job_output_template)" -D $exehome $(srunargs) $exename $exeflags $(worker_arg())`
srun_proc = open(srun_cmd)
slurm_spec_regex = r"([\w]+):([\d]+)#(\d{1,3}.\d{1,3}.\d{1,3}.\d{1,3})"
retry_delays = manager.retry_delays
for i = 0:np - 1
println("connecting to worker $(i + 1) out of $np")
slurm_spec_match = nothing
fn = make_job_output_path(lpad(i, 4, "0"))
t0 = time()
while true
for retry_delay in retry_delays
# Wait for output log to be created and populated, then parse
if isfile(fn) && filesize(fn) > 0
slurm_spec_match = open(fn) do f
Expand All @@ -77,8 +83,13 @@ function launch(manager::SlurmManager, params::Dict, instances_arr::Array,
end
end
# Sleep for some time to limit ressource usage while waiting for the job to start
sleep(0.1)
sleep(retry_delay)
end

if slurm_spec_match === nothing
throw(SlurmException("Timeout while trying to connect to worker"))
end

config = WorkerConfig()
config.port = parse(Int, slurm_spec_match[2])
config.host = strip(slurm_spec_match[3])
Expand All @@ -99,4 +110,21 @@ function manage(manager::SlurmManager, id::Integer, config::WorkerConfig,
# This function needs to exist, but so far we don't do anything
end

addprocs_slurm(np::Integer; kwargs...) = addprocs(SlurmManager(np); kwargs...)
SlurmManager(np::Integer) = SlurmManager(np, ExponentialBackOff(n=10, first_delay=1,
max_delay=512, factor=2))

"""
Launch `np` workers on a cluster managed by slurm. `retry_delays` is a vector of
numbers specifying in seconds how long to repeatedly wait for a worker to start.
Defaults to an exponential backoff.

# Examples

```
addprocs_slurm(100; retry_delays=Iterators.repeated(0.1))
```
"""
addprocs_slurm(np::Integer;
retry_delays=ExponentialBackOff(n=10, first_delay=1,
max_delay=512, factor=2),
kwargs...) = addprocs(SlurmManager(np, retry_delays); kwargs...)