Seg fault while trying to ingest

[stockholmux]

While attempting to run the benchmark, I keep running into seg faults.

I'm running Redis 4.02 (OSS) / RediSearch 0.21.3 with 5 nodes. While running the benchmark using this:

./RediSearchBenchmark -engine redis -shards 5 -hosts "localhost:6379,localhost:6380,localhost:6381,localhost:6382,localhost:6383"     -file ~/enwiki-latest-abstract.xml

I'm guessing this is a RediSearch issue, so I'll x-post an issue over on that repo.

Here are the dumps - I was pretty consistently getting 2 out of 5 shards seg faulting.

redisearch-ingest-seg-fault2.txt
redisearch-ingest-seg-fault1.txt

[dvirsky]

There is a bug in redis 4.0.2 that was fixed already, run it with redis unstable, it shouldn't happen.

[stockholmux]

Gotcha. I'll run it and report back.

[dvirsky]

It might be a different issue, since the benchmark doesn't do multi IIRC. But using the latest redis is a must in this context.

…

On Fri, Oct 20, 2017 at 6:31 PM Kyle J. Davis ***@***.***> wrote: Gotcha. I'll run it and report back. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAetupmHRI42ewUNmMT0y1wVQUBySL7cks5suLzigaJpZM4QABqr> .

[stockholmux]

Still getting seg faults in a similar pattern (run the benchmark, within a few seconds two nodes give segs, the rest run with no problem) with unstable.

Not sure if this is relevant, but I'm running the bog standard install of unstable with redis.io provided conf file and redisearch and port set via the command line when starting the servers.

seg-fault-4.txt
seg-fault-3.txt

I may try rolling back the versions of RediSearch until I hit one that works with the benchmark.

[stockholmux]

This must be a relatively recent bug. Unstable + RediSearch 0.20.0 and 0.21.0 do not cause seg faults.

[dvirsky]

ok, I'll try to recreate. What are you using to run the benchmark? wikipedia extracts?

…

On Mon, Oct 23, 2017 at 2:31 PM Kyle J. Davis ***@***.***> wrote: This must be a relatively recent bug. Unstable + RediSearch 0.20.0 and 0.21.0 do not cause seg faults. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAeturwui2WPan3yC8_3mAhiAPexhdJkks5svJU9gaJpZM4QABqr> .

[stockholmux]

Yup. Wikipedia extracts.

[dvirsky]

@mnunberg do you mind having a look?

[mnunberg]

ok, I seem to have triggered it.

[mnunberg]

@stockholmux btw did you get this on Mac or Linux?

[stockholmux]

@mnunberg AWS / Ubuntu: c4.4xlarge.

[mnunberg]

I believe I've fixed the issue. Can you verify? (latest master on RediSearch)

[stockholmux]

@mnunberg I won't have a chance until Thursday, unfortunately.

[filipecosta90]

closing given this issue is no longer relevant/applicable.