SharedArray Bus error #15155
Closed
jltmtz opened this Issue on Feb 19 · 8 comments
Projects
None yet
Labels
needs more info
parallel
Milestone
No milestone
Assignees
No one assigned
7 participants
@jltmtz
@tkelman
@ViralBShah
@JaredCrean2
@skariel
@amitmurthy
@vtjnash
Notifications

You’re not receiving notifications from this thread.
@jltmtz
jltmtz commented on Feb 19

I am running a program that spawns several local processes and then gathers the results back. It works fine most of the time, but if I let it run for around a day, I eventually get this error:

signal (7): Bus error
setindex! at sharedarray.jl:285
jlcall_setindex!_22036 at  (unknown line)
jl_apply_generic at /ubc/cs/research/tracking-raid/julm/libraries/julia-4.0/julia-e5c6964a49/bin/../lib/julia/libjulia.so (unknown line)

(more stuff from my own code)

Like I said, the error is somewhat erratic in that this piece of code runs many many times and then suddenly fails.
@tkelman
The Julia Language member
tkelman commented on Feb 19

Can you point us to something reproducible that we would be able to run? Without that I don't think there's enough to go on here yet.
@ViralBShah
The Julia Language member
ViralBShah commented on Feb 21

Cc @amitmurthy
@ViralBShah
The Julia Language member
ViralBShah commented on Feb 21

Definitely need a reproducible test case.
@ViralBShah ViralBShah added the parallel label on Feb 21
@JaredCrean2
JaredCrean2 commented on Feb 21

I have seen the same thing running a serial code for roughly 9 hours, but I haven't been able to reproduce it with any consistency.
@skariel
skariel commented on Feb 23

I have a similar thing, I think it happens only when I use very large arrays. Using smaller ones with same code everything is OK.
@jltmtz
jltmtz commented on Feb 23

Sorry, I tried but I wasn't able to reproduce.

Should I close this issue or leave it open for others/future reference? Currently google returns nothing when one googles this error.
@amitmurthy
The Julia Language member
amitmurthy commented on Feb 23

Since other folks have also reported the same issue, lets leave it open for now.
@vtjnash vtjnash added the needs-more-info label on Mar 8
This was referenced on Mar 9
Merged
use of finalizers to release remote references/shared arrays early. #15436
Open
Large number of temporary shared arrays crashes Julia. Need a more aggressive gc? #15415
@amitmurthy
The Julia Language member
amitmurthy commented on Mar 13

This issue appears to the be the same as #15415 . Closing it for now. If on 0.5 please try with calling finalizers to force an early release. If on 0.4 you will need #15441 merged to do the same.

Please reopen if you have a reproducible test case different from #15415
@amitmurthy amitmurthy closed this on Mar 13