memory issue with v3.3
I realize v3.3 is now three versions old, but thought I would put this on the record anyway.
I was invoking our HPC's installed AUGUSTUS 3.3 from a BUSCO config file. Kept getting core dumps and quite bad (incomplete) BUSCO results. As an alternate I tried using the downloadable BUSCO VM and it worked perfectly; the AUGUSTUS version in that is 3.2.2. I reported all this to our HPC team. Investigation of the faulty process showed this:
Mar 26 13:03:33 c33-01 kernel: memory: usage 41943040kB, limit 41943040kB, failcnt 9376
Mar 26 13:03:33 c33-01 kernel: memory+swap: usage 41943040kB, limit 41943040kB, failcnt 0
Mar 26 13:03:33 c33-01 kernel: kmem: usage 0kB, limit 9007199254740988kB, failcnt 0
Mar 26 13:03:33 c33-01 kernel: Memory cgroup stats for /slurm/uid_1415749/job_347628/step_batch: cache:0KB rss:41943040KB rss_huge:0KB mapped_file:0KB swap:0KB inactive_anon:436608KB active_anon:4150
6392KB inactive_file:0KB active_file:0KB unevictable:0KB
Mar 26 13:03:33 c33-01 kernel: [ pid ] uid tgid total_vm rss nr_ptes swapents oom_score_adj name
Mar 26 13:03:33 c33-01 kernel: [27782] 1415749 27782 28366 452 9 0 0 slurm_script
Mar 26 13:03:33 c33-01 kernel: [27800] 1415749 27800 73914 3631 60 0 0 python
Mar 26 13:03:33 c33-01 kernel: [28266] 1415749 28266 10509235 10483962 20511 0 0 augustus
Mar 26 13:03:33 c33-01 kernel: Memory cgroup out of memory: Kill process 28266 (augustus) score 1001 or sacrifice child
Mar 26 13:03:33 c33-01 kernel: Killed process 28266 (augustus) total-vm:42036940kB, anon-rss:41933020kB, file-rss:2828kB, shmem-rss:0kB
Our IT maven said the v3.3 problem was a 'memory leak' "It turned out augustus 3.3 does not work well, memory usage continues increase till run out of memory"
So he built me a singularity package of BUSCO (using AUGUSTUS 3.2.2) and it works as it should. No core dumps, good BUSCO results.
I haven't yet tried the newer 3.3.1/3.3.2 versions of AUGUSTUS for comparison.
The same issue persists on 3.5.0