I am facing an issue regarding sort being killed due over usage of memory and can not understand why it surpasses the given limit by a significant margin.
Context:
Available physical RAM: 8GB
Memory in use by system: <500MB (FreeBSD 13)
Free memory: ~7.5GB
Swap: 2G
No system hard limit on mem usage ($ ulimit -> unlimited)
File being sorted: 300+ GB
HDD space: 1Tb+ free both on target or temporary dir.
Even when I run sort specifying a resource limit whithin the available; or without any explicit argument (which should automatically be set to 90% of available) it eventually gets killed due over usage.
[mario@casa-server ~/pjy]$ sort -S 7G -T /filesWD-WMAY03648941/tmp/ -o /filesWD-WMAY03648941/pjA-sorted.txt pjyA.txt
Killed
[mario@casa-server ~/pjy]$
[mario@casa-server ~/pjy]$ sort -T /filesWD-WMAY03648941/tmp/ -o /filesWD-WMAY03648941/pjA-sorted.txt pjyA.txt
Killed
[mario@casa-server ~/pjy]$
Apparently it is indeed over using the resource limit. When run with -S 3G I can observe around 5G of usage sometimes.
# ~5G USAGE EVEN WITH EXPLICIT -S 3G LIMIT
[mario@casa-server ~]$ sudo procstat -r 3183
PID COMM RESOURCE VALUE
3183 sort user time 00:27:51.849171
3183 sort system time 00:03:28.220262
3183 sort maximum RSS 5423072 KB
3183 sort integral shared memory 10505728 KB
3183 sort integral unshared data 1910064 KB
3183 sort integral unshared stack 30560128 KB
3183 sort page reclaims 8172561
3183 sort page faults 1724
3183 sort swaps 0
3183 sort block reads 24296
3183 sort block writes 17719
3183 sort messages sent 0
3183 sort messages received 0
3183 sort signals received 0
3183 sort voluntary context switches 5868
3183 sort involuntary context switches 30700
[mario@casa-server ~]$
# TOP excerpt
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
3183 mario 1 103 0 4657M 4603M CPU3 3 38:23 99.88% sort -S 3G -T /filesWD-WMAY03648941/tmp/ -o /files
# NO SO LIMIT ON USAGE
[mario@casa-server ~]$ sudo procstat -l 3183
PID COMM RLIMIT SOFT HARD
3183 sort cputime infinity infinity
3183 sort filesize infinity infinity
3183 sort datasize 32768 MB 32768 MB
3183 sort stacksize 524288 KB 524288 KB
3183 sort coredumpsize infinity infinity
3183 sort memoryuse infinity infinity
3183 sort memorylocked 65536 B 65536 B
3183 sort maxprocesses 12186 12186
3183 sort openfiles 234576 234576
3183 sort sbsize infinity infinity
3183 sort vmemoryuse infinity infinity
3183 sort pseudo-terminals infinity infinity
3183 sort swapuse infinity infinity
3183 sort kqueues infinity infinity
3183 sort umtxp infinity infinity
[mario@casa-server ~]$
[mario@casa-server ~]$ ulimit
unlimited
That explain why it is being killed when ran with 7G limit (of 7.5 available), or when the limit was automatically set: it was indeed over using all the mem+swap (and being killed by OS as expected).
But what am I missing here? man sort:
-S size, --buffer-size=size
Use size for the maximum size of the memory buffer. Size
modifiers %,b,K,M,G,T,P,E,Z,Y can be used. If a memory limit is
not explicitly specified, sort takes up to about 90% of available
memory. If the file size is too big to fit into the memory
buffer, the temporary disk files are used to perform the sorting.
If -S 3G is set, why is it over using RAM and not the HDD temporary folder?
PS: I tried to ask this question related to other 'sort being killed due out of memory' but it was mismanaged by some moderators. I do think it is a related question (and flagging it as a unrelated one just fragments the solution). Also, I read through many other possible solutions and they are not suited to this case.
Thanks all
-T?