ClusterCockpit/cc-condor-sync
1
0
Fork 0
mirror of https://gitlab.cs.uni-saarland.de/hpc/cc-condor-sync.git synced 2025-10-30 08:45:07 +01:00
Code Issues Packages Projects Releases Wiki Activity

Make stop script more robust.

Browse Source 
And handle pagination.
This commit is contained in:
Joachim Meyer
2023-05-08 09:40:50 +02:00
parent 89f6440e0f
commit 7a0c41e378
1 changed files with 18 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
scripts/stop-jobs-no-longer-running.py
View File

@@ -35,12 +35,21 @@ class CCApi:
def getJobs(self, filter_running=True): def getJobs(self, filter_running=True):
url = self.apiurl+"jobs/" url = self.apiurl+"jobs/"
if filter_running: if filter_running:
url = url+"?state=running" url = url+"?state=running&items-per-page=100"
r = requests.get(url, headers=self.headers) jobs = []
if r.status_code == 200: page = 1
return r.json()['jobs'] while True:
else: r = requests.get(url + "&page=%d" % page, headers=self.headers)
return [] page += 1
if r.status_code == 200:
new_jobs = r.json()['jobs']
if len(new_jobs) < 100:
jobs.extend(new_jobs)
return jobs
else:
jobs.extend(new_jobs)
else:
return []
def _getSubmitNodeId(self, globalJobId): def _getSubmitNodeId(self, globalJobId):
job_id_parts = globalJobId.split('#') job_id_parts = globalJobId.split('#')
@@ -95,11 +104,12 @@ if __name__ == "__main__":
cc = CCApi(config, args.debug) cc = CCApi(config, args.debug)
condor_jobs = subprocess.run( condor_jobs = subprocess.run(
["ssh", "conduit2.cs.uni-saarland.de", "condor_q", "-json", "-all", "-glob"], capture_output=True, text=True).stdout ["ssh", "conduit2.cs.uni-saarland.de", "condor_q", "-json", "-all", "-glob", "-constraint", "\"JobStatus == 2\""], capture_output=True, text=True).stdout
running_jobs = json.loads(condor_jobs) running_jobs = json.loads(condor_jobs)
cc_jobs = cc.getJobs() cc_jobs = cc.getJobs()
running_job_dict = {cc._jobIdToInt(job['GlobalJobId']): (job['GlobalJobId'], job['EnteredCurrentStatus']) for job in running_jobs} running_job_dict = {cc._jobIdToInt(job['GlobalJobId']): (job['GlobalJobId'], job['JobCurrentStartDate']) for job in running_jobs}
print("CC jobs:", len(cc_jobs), " Condor jobs:", len(running_job_dict))
for cc_job in cc_jobs: for cc_job in cc_jobs:
startTime = cc_job['startTime'] startTime = cc_job['startTime']
if not cc_job['jobId'] in running_job_dict or abs(startTime - running_job_dict[cc_job['jobId']][1]) > 5: if not cc_job['jobId'] in running_job_dict or abs(startTime - running_job_dict[cc_job['jobId']][1]) > 5: