Hello,
We're seeing job failures because users have been setting up job
dependencies where one job depends on many (thousands), and we hit the
1024 string limit, partly because job ids are 7 digits right now. The
following patch increases the limit, but opens up the possibility that
"malicious" users might use up too much memory at the controller.
I've also increased the limit for job name a little bit, as we've hit
that limit as well.
I'm not sure this is the right way to handle this, but thought I would
submit the one-line patch in-case others are seeing the same thing.
-sam
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 891e5a0..f442101 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -4078,7 +4078,7 @@ static int
_validate_job_create_req(job_desc_msg_t * job_desc)
_test_strlen(job_desc->ckpt_dir, "ckpt_dir", 1024) ||
_test_strlen(job_desc->comment, "comment", 1024) ||
_test_strlen(job_desc->cpu_bind, "cpu_bind", 1024) ||
- _test_strlen(job_desc->dependency, "dependency", 1024) ||
+ _test_strlen(job_desc->dependency, "dependency", 1024*512) ||
_test_strlen(job_desc->exc_nodes, "exc_nodes", 1024*64) ||
_test_strlen(job_desc->features, "features", 1024) ||
_test_strlen(job_desc->gres, "gres", 1024) ||
@@ -4087,7 +4087,7 @@ static int
_validate_job_create_req(job_desc_msg_t * job_desc)
_test_strlen(job_desc->mail_user, "mail_user", 1024) ||
_test_strlen(job_desc->mem_bind, "mem_bind", 1024) ||
_test_strlen(job_desc->mloaderimage, "mloaderimage", 1024) ||
- _test_strlen(job_desc->name, "name", 1024) ||
+ _test_strlen(job_desc->name, "name", 1024*16)
||
_test_strlen(job_desc->network, "network", 1024) ||
_test_strlen(job_desc->partition, "partition", 1024) ||
_test_strlen(job_desc->qos, "qos", 1024) ||
