[rancid] Problem with hpuifilter on RHEL6 and glibc-2.12-1.47.el6.x86_64

[Andy Cobaugh]

Hey folks.

We recently saw a problem crop up with hpuifilter on RHEL6 (well,
Scientific Linux 6.1) when glibc/glibc-common got updated.

We're running RANCID 2.3.6 with git extensions. The problem only happens
with HP switches, so I'm thinking this is a problem in hpuifilter
somewhere, but maybe it's in tcl/expect/ssh.

The problem exists with glibc-2.12-1.47.el6.x86_64, but not with
glibc-2.12-1.25.el6.x86_64, which came from 6.1 FCS. I'm not sure what
version we were running before the update.

The diff outputs we're seeing look something like this:

-aaa port-access authenticator 8 client-limit 1
-aaa port-access authenticator 9 quiet-period 30
+aaa port-access authenticator 8 client-laaa poaaa port-access authenticator 9 quiet-period 30

-;Image: stamp: /sw/code/build/cod(cod11)
+;Image: ^[[24magemage stamp: /sw/code/build/cod(cod11)

-ip default-gatew.1.1
-sntp sesntp server 128.118.25.3
+ip default-gateway 10.1.1.1
+sntp server 128.118.25.3


It's completely random which switches show this behavior, but all of them
tend to flip-flop between ok and not ok, and we have enough switches that
we get an email every time RANCID runs.

I'm not sure where else to look. Open to suggestions.

--andy

[heasley]

Post by Andy Cobaugh
Hey folks.
We recently saw a problem crop up with hpuifilter on RHEL6 (well,
Scientific Linux 6.1) when glibc/glibc-common got updated.
We're running RANCID 2.3.6 with git extensions. The problem only happens
with HP switches, so I'm thinking this is a problem in hpuifilter
somewhere, but maybe it's in tcl/expect/ssh.
The problem exists with glibc-2.12-1.47.el6.x86_64, but not with
glibc-2.12-1.25.el6.x86_64, which came from 6.1 FCS. I'm not sure what
version we were running before the update.
-aaa port-access authenticator 8 client-limit 1
-aaa port-access authenticator 9 quiet-period 30
+aaa port-access authenticator 8 client-laaa poaaa port-access authenticator 9 quiet-period 30
-;Image: stamp: /sw/code/build/cod(cod11)
+;Image: ^[[24magemage stamp: /sw/code/build/cod(cod11)
-ip default-gatew.1.1
-sntp sesntp server 128.118.25.3
+ip default-gateway 10.1.1.1
+sntp server 128.118.25.3
It's completely random which switches show this behavior, but all of them
tend to flip-flop between ok and not ok, and we have enough switches that
we get an email every time RANCID runs.
I'm not sure where else to look. Open to suggestions.

are you sure that your have rancid 2.3.6? hpuifilter had been patched to
avoid some internationalized string functions which didnt like the 8 bit
chars of the screen handling codes. perhaps something else has contracted
this disease, but first please check that you really have 2.3.6.

[Andy Cobaugh]

Post by heasley
are you sure that your have rancid 2.3.6? hpuifilter had been patched to
avoid some internationalized string functions which didnt like the 8 bit
chars of the screen handling codes. perhaps something else has contracted
this disease, but first please check that you really have 2.3.6.

That's what I thought at first, and we were running 2.3.3 before, so I
upgraded to 2.3.6 and saw the same behavior.

And indeed, we are running 2.3.6 in production:

$ /afs/bx.psu.edu/service/rancid/prod/bin/hpuifilter -v
hpuifilter: rancid version 2.3.6

$ /afs/bx.psu.edu/service/rancid/prod/bin/rancid-run -V
rancid 2.3.6

--andy

[Andy Cobaugh]

Post by heasley

Post by Andy Cobaugh
Hey folks.
We recently saw a problem crop up with hpuifilter on RHEL6 (well,
Scientific Linux 6.1) when glibc/glibc-common got updated.
We're running RANCID 2.3.6 with git extensions. The problem only happens
with HP switches, so I'm thinking this is a problem in hpuifilter
somewhere, but maybe it's in tcl/expect/ssh.
The problem exists with glibc-2.12-1.47.el6.x86_64, but not with
glibc-2.12-1.25.el6.x86_64, which came from 6.1 FCS. I'm not sure what
version we were running before the update.
-aaa port-access authenticator 8 client-limit 1
-aaa port-access authenticator 9 quiet-period 30
+aaa port-access authenticator 8 client-laaa poaaa port-access authenticator 9 quiet-period 30
-;Image: stamp: /sw/code/build/cod(cod11)
+;Image: ^[[24magemage stamp: /sw/code/build/cod(cod11)
-ip default-gatew.1.1
-sntp sesntp server 128.118.25.3
+ip default-gateway 10.1.1.1
+sntp server 128.118.25.3
It's completely random which switches show this behavior, but all of them
tend to flip-flop between ok and not ok, and we have enough switches that
we get an email every time RANCID runs.
I'm not sure where else to look. Open to suggestions.

are you sure that your have rancid 2.3.6? hpuifilter had been patched to
avoid some internationalized string functions which didnt like the 8 bit
chars of the screen handling codes. perhaps something else has contracted
this disease, but first please check that you really have 2.3.6.

Ping.

Anyone have a chance to look at this? We are definitely running the latest
rancid, but the problem only exists with the newer glibc found on RHEL6.
Older versions from a few months ago are fine.

--andy