You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A job's "Variable_List" attribute appears to get dropped if an attempt at job submission does not go through on the first try. This is likely to not be isolated to qsub, but that is the only way I have attempted to demonstrate it.
qsub makes multiple attempts at job submission in cmds/qsub_functions.c main_func(). I tripped upon this particular issue when submitting a job with a dependency that sent back a PBSE_BADDEPEND on first try but then succeeded once alternate_dependency was substituted in.
My amateur digging has led me to src/lib/Libifl/env_attropl_hash.c build_var_list(), where I can see that ATTR_v is deleted from the hash. It's not 100% clear why this is done, other than I noticed that the ATTR_v is sent last in encode_DIS_attropl_hash() and dropping ATTR_v would prevent it from being sent twice. If that is the case a simple if check in enc_DIS_attropl_hash_single() might be better than deleting from the hash table and losing state in between submission attempts.
The text was updated successfully, but these errors were encountered:
A job's "Variable_List" attribute appears to get dropped if an attempt at job submission does not go through on the first try. This is likely to not be isolated to qsub, but that is the only way I have attempted to demonstrate it.
qsub makes multiple attempts at job submission in cmds/qsub_functions.c main_func(). I tripped upon this particular issue when submitting a job with a dependency that sent back a PBSE_BADDEPEND on first try but then succeeded once alternate_dependency was substituted in.
My amateur digging has led me to src/lib/Libifl/env_attropl_hash.c build_var_list(), where I can see that ATTR_v is deleted from the hash. It's not 100% clear why this is done, other than I noticed that the ATTR_v is sent last in encode_DIS_attropl_hash() and dropping ATTR_v would prevent it from being sent twice. If that is the case a simple if check in enc_DIS_attropl_hash_single() might be better than deleting from the hash table and losing state in between submission attempts.
The text was updated successfully, but these errors were encountered: