[Bug 260438] dns/bind-tools: dig SIGABRT under high load
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 15 Dec 2021 14:24:31 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=260438
Bug ID: 260438
Summary: dns/bind-tools: dig SIGABRT under high load
Product: Ports & Packages
Version: Latest
Hardware: Any
OS: Any
Status: New
Severity: Affects Some People
Priority: ---
Component: Individual Port(s)
Assignee: mat@FreeBSD.org
Reporter: david@isnic.is
Flags: maintainer-feedback?(mat@FreeBSD.org)
Assignee: mat@FreeBSD.org
We have some shell scripts to set up zones on new slaves. Since we have a large
number (tens of thousands) of zones, we do this in parallel, querying the slave
to see if the zone is set up and if not send an `rndc addzone`.
When doing a bulk provisioning like this, we see dig very occasionally die with
SIGABRT:
```
[root@hfp-master /usr/home/ansible]# dmesg
pid 72286 (dig), jid 0, uid 0: exited on signal 6 (core dumped)
```
Error ouput when this happens:
```
dighost.c:2628: REQUIRE((__builtin_expect(!!(((query)) != ((void *)0)), 1) &&
__builtin_expect(!!(((const isc__magic_t *)((query)))->magic == ((('D') << 24 |
('i') << 16 | ('g') << 8 | ('q')))), 1))) failed, back trace
#0 0x4359ba in ??
#1 0x43594a in ??
#2 0x2bc814 in ??
#3 0x462a33 in ??
#4 0x44ade8 in ??
#5 0x447225 in ??
#6 0x800a39ada in ??
#7 0x800a4ac1b in ??
#8 0x800a3a051 in ??
#9 0x44735b in ??
#10 0x465135 in ??
Abort trap (core dumped)
```
This only seems to happen under load, running multiple dig commands in parallel
in a tight loop.
I've tried to create a concise repro case here without sharing our whole DNS
deployment script set, but don't have it ready. I'll add it later if I get it
working.
This looks extremely similar to the following upstream bugs:
https://gitlab.isc.org/isc-projects/bind9/-/issues/1981
https://gitlab.isc.org/isc-projects/bind9/-/issues/1971
https://gitlab.isc.org/isc-projects/bind9/-/issues/1956
Based on comments there, this should be fixed in this MR:
https://gitlab.isc.org/isc-projects/bind9/-/merge_requests/3721
However it seems this still happens on FreeBSD.
We are running bind-tools-9.16.23.
--
You are receiving this mail because:
You are the assignee for the bug.