SIGSEGV in Jetty

[Phil H <gi...@gmail.com>]

Hi there,

I am getting regular (maybe every ten minutes?) crashes in NiFi 1.3.0. This
just started happening unrelated to any change to the software environment
(i.e.: we haven't installed new code/processors).

It roughly coincides with an increase in flow file throughout and my
subsequent use of multiple concurrent tasks on some processors. However I
set those back to 1, and the problem has persisted.

I can't get the error file off the customer network, but the problematic
frame is:

org.eclipse.jetty.io.IdleTimeout.deactivate()V+8

Is this a known issue, and if so what are my options?

TIA,
Phil

Re: SIGSEGV in Jetty

[Phil H <gi...@gmail.com>]

Hi Joe,

Early days, but that change appears to have fixed the issue. I note it
occurred sooner if I used the web UI a lot (for example if I was
configuring processors or checking queues).

Thanks again

Phil.

On Thu, 26 Apr 2018 at 12:47, Joe Witt <jo...@gmail.com> wrote:

> Phil,
>
> I'd definitely say not any kind of known issue.  One thing you might
> want to check is whether you're using G1 GC. We have in most cases
> moved away from that in Java8 because of some long standing bugs that
> our Lucene indexing for our provenance data which could trigger it.
> We've switch to using the default GC in many cases.
>
> In your nifi conf/bootstrap.conf look for
>
> java.arg.13=-XX:+UseG1GC
>
> comment that line out by doing
>
> #java.arg.13=-XX:+UseG1GC
>
> And see if that helps.
>
> Thanks
>
> On Wed, Apr 25, 2018 at 10:43 PM, Phil H <gi...@gmail.com> wrote:
> > Hi there,
> >
> > I am getting regular (maybe every ten minutes?) crashes in NiFi 1.3.0.
> This
> > just started happening unrelated to any change to the software
> environment
> > (i.e.: we haven't installed new code/processors).
> >
> > It roughly coincides with an increase in flow file throughout and my
> > subsequent use of multiple concurrent tasks on some processors. However I
> > set those back to 1, and the problem has persisted.
> >
> > I can't get the error file off the customer network, but the problematic
> > frame is:
> >
> > org.eclipse.jetty.io.IdleTimeout.deactivate()V+8
> >
> > Is this a known issue, and if so what are my options?
> >
> > TIA,
> > Phil
>

Re: SIGSEGV in Jetty

[Joe Witt <jo...@gmail.com>]

Phil,

I'd definitely say not any kind of known issue.  One thing you might
want to check is whether you're using G1 GC. We have in most cases
moved away from that in Java8 because of some long standing bugs that
our Lucene indexing for our provenance data which could trigger it.
We've switch to using the default GC in many cases.

In your nifi conf/bootstrap.conf look for

java.arg.13=-XX:+UseG1GC

comment that line out by doing

#java.arg.13=-XX:+UseG1GC

And see if that helps.

Thanks

On Wed, Apr 25, 2018 at 10:43 PM, Phil H <gi...@gmail.com> wrote:
> Hi there,
>
> I am getting regular (maybe every ten minutes?) crashes in NiFi 1.3.0. This
> just started happening unrelated to any change to the software environment
> (i.e.: we haven't installed new code/processors).
>
> It roughly coincides with an increase in flow file throughout and my
> subsequent use of multiple concurrent tasks on some processors. However I
> set those back to 1, and the problem has persisted.
>
> I can't get the error file off the customer network, but the problematic
> frame is:
>
> org.eclipse.jetty.io.IdleTimeout.deactivate()V+8
>
> Is this a known issue, and if so what are my options?
>
> TIA,
> Phil