[bind10-dev] systest failures

Jeremy C. Reed jreed at isc.org
Wed Jun 22 21:00:55 UTC 2011


I am using make systest on FreeBSD 8.2-RELEASE. I have 
BIND9_TOP=/usr/home/jreed//bind9/bind-9.8.0 in my environment.

I have this patch: http://bind10.isc.org/~jreed/systest.diff.txt for 
tests/system/start.pl so I can set the msqg-socket due to bug #360 and 
to show a little more debugging output.

This resulted in:

cd tests/system;  sh /usr/home/jreed/builder/work/BIND10-cppcheck/20110621173001-FreeBSD8-amd64-GCC/build/tests/system/runall.sh
S:bindctl:Wed Jun 22 14:58:14 UTC 2011
T:bindctl:1:A
A:System test bindctl
I:check ports
I:starting server nsx1
I:verify server
I:Checking b10-auth is working by default (0)
I:Checking BIND 10 statistics after a pause (1)
I:Stopping b10-auth and checking that (2)

; <<>> DiG 9.8.0 <<>> +tcp +norec @10.53.0.1 -p 53210 ns.example.com. A
; (1 server found)
;; global options: +cmd
;; connection timed out; no servers could be reached

**** there was a delay for maybe 5 minutes here ****

I:Restarting b10-auth and checking that (3)
I:failed
I:Rechecking BIND 10 statistics after a pause (4)
I:failed
I:Changing the data source from sqlite3 to in-memory (5)
 
I:failed
I:Rechecking BIND 10 statistics after changing the datasource (6)
I:failed
I:exit status: 1
R:FAIL
E:bindctl:Wed Jun 22 15:00:13 UTC 2011
S:glue:Wed Jun 22 15:00:13 UTC 2011
T:glue:1:A
A:System test glue
I:check ports
I:starting server nsx1
I:verify server
I:testing that a TLD referral gets a full glue set from the root zone 
(0)
I:testing that we find glue A RRs we are authoritative for (1)
I:testing that we don't find out-of-zone glue (2)
I:exit status: 0
R:PASS
E:glue:Wed Jun 22 15:00:45 UTC 2011
*** Error code 1

Stop in 
/usr/home/jreed/builder/work/BIND10-cppcheck/20110621173001-FreeBSD8-amd64-GCC/build.


Then I ran it another time. Different install, same patch, not already 
running.

cd tests/system;  sh 
/usr/home/jreed/builder/work/BIND10-cppcheck/20110622153013-FreeBSD8-amd64-GCC/build/tests/system/runall.sh
S:bindctl:Wed Jun 22 19:01:15 UTC 2011
T:bindctl:1:A
A:System test bindctl
I:check ports
I:starting server nsx1
I:Couldn't start server nsx1
R:FAIL
S:glue:Wed Jun 22 19:01:49 UTC 2011
T:glue:1:A
A:System test glue
I:check ports
/usr/home/jreed//bind9/bind-9.8.0/bin/tests/system/testsock.pl: 
bind(10.53.0.1, 53210): Address already in use
I:Couldn't bind to socket (yet)
/usr/home/jreed//bind9/bind-9.8.0/bin/tests/system/testsock.pl: 
bind(10.53.0.1, 53210): Address already in use
I:Couldn't bind to socket (yet)
/usr/home/jreed//bind9/bind-9.8.0/bin/tests/system/testsock.pl: 
bind(10.53.0.1, 53210): Address already in use
I:Couldn't bind to socket (yet)
/usr/home/jreed//bind9/bind-9.8.0/bin/tests/system/testsock.pl: 
bind(10.53.0.1, 53210): Address already in use
I:Couldn't bind to socket (yet)
/usr/home/jreed//bind9/bind-9.8.0/bin/tests/system/testsock.pl: 
bind(10.53.0.1, 53210): Address already in use
start.pl: could not bind to server addresses, still running?
I:server sockets not available
R:FAIL

I killed that with Ctrl-C.

I saved the run output at http://bind10.isc.org/~jreed/bind10.run.2.txt

I don't know why "Address already in use"; I assume because the bindctl 
test did start and it didn't recognize it so it didn't stop it.

I already know about the xfrout socket issue. I sent a long email about 
environment variables about this some weeks ago. run_bind10.sh sets some 
environment variables that cause it to override other environment 
variables. I will make patches for this, but the problem with too many 
ways to use and too many environment variables is a bigger problem.

But even with that problem b10-auth should work.

Another problem see end of that output shows SIGKILLs. I will open a 
ticket for that. That should not happen.

So I make sure BIND 10 is not running and I run systest again. Different 
output a third time:

cd tests/system;  sh 
/usr/home/jreed/builder/work/BIND10-cppcheck/20110622153013-FreeBSD8-amd64-GCC/build/tests/system/runall.sh
S:bindctl:Wed Jun 22 19:03:13 UTC 2011
T:bindctl:1:A
A:System test bindctl
I:check ports
I:starting server nsx1
I:verify server
I:Checking b10-auth is working by default (0)
I:Checking BIND 10 statistics after a pause (1)
I:Stopping b10-auth and checking that (2)
;; Connection to 10.53.0.1#53210(10.53.0.1) for ns.example.com. failed: 
connection refused.
I:Restarting b10-auth and checking that (3)
I:Rechecking BIND 10 statistics after a pause (4)
I:Changing the data source from sqlite3 to in-memory (5)
I:Rechecking BIND 10 statistics after changing the datasource (6)
I:exit status: 0
R:PASS
E:bindctl:Wed Jun 22 19:03:43 UTC 2011
S:glue:Wed Jun 22 19:03:43 UTC 2011
T:glue:1:A
A:System test glue
I:check ports
I:starting server nsx1
I:verify server
I:testing that a TLD referral gets a full glue set from the root zone 
(0)
I:testing that we find glue A RRs we are authoritative for (1)
I:testing that we don't find out-of-zone glue (2)
I:exit status: 0
R:PASS
E:glue:Wed Jun 22 19:04:02 UTC 2011

(BIND 10 was all exited then.)

The output is at 
http://bind10.isc.org/~jreed/bind10.run.3.txt
(for bindctl)
and
http://bind10.isc.org/~jreed/glue-bind10.run.3.txt
(for the glue test)


I do not understand yet why I had three different results from running 
this.

Also where is the S, T, A, I, R, E documented ?

I think:
 S is start time?
 R is result?
 E is end time?
 what about the others?

For what it is worth, I was able to run BIND 10 manually and it worked 
okay. (My long email yesterday about all the strange hot cache and 
related messages was same system.)



More information about the bind10-dev mailing list