[j-nsp] EX-Series experiences

[Sven Juergensen (KielNET)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi list,

once again I had a strange occurrence with
one EX420048T switch running JUNOS 9.3R2.8.

Several VLANs and layer3 routing are confi-
gured on it. The symptom was that directly
connected (access-) VLANs x and y couldn't
talk to an address range, which is routed
through a trunk to VLAN x's IP on another
switch.

Mind you, there is nothing fancy configured
on the switches and they're not bogged down
to their knees. Perhaps it's an interoper-
ability problem but I wouldn't know that.
All this is basic functionality that tends
to go haywire though, preferably when people
are busy not being awake.

Things happened without any human intervention
or after recent configuration changes. This
time, the logs actually provided something
which I can't make heads or tails from the
logs so here goes:

- ---8<---
Jan 30 22:26:33  some.switch fpc0 Resolve request came for an address
matching on Wrong nh nh:1317, type:Unicast...?
Jan 30 22:26:45  some.switch last message repeated 8 times
Jan 30 22:28:45  some.switch last message repeated 167 times
Jan 30 22:38:46  some.switch last message repeated 913 times
Jan 30 22:48:44  some.switch last message repeated 713 times
Jan 30 22:55:02  some.switch last message repeated 262 times
Jan 30 22:55:30  some.switch fpc0 Resolve request came for an address
matching on Wrong nh nh:1317, type:Unicast...?
Jan 30 22:55:45  some.switch last message repeated 7 times
Jan 30 22:57:46  some.switch last message repeated 179 times
- --->8---

The MAC and ARP table at that time looked
just fine and I doubt it's related to the
spanning-tree, since so far it's a dual
star topology without any loops.

Neither restarting the switching nor rou-
ting daemon helped anything; a complete
reboot of the switch did the trick.

Perhaps this is of interest to Juniper, I
don't really expect any help with it. Maybe
one of the future JUNOS releases will
remedy that.

Nonetheless, I'm kind of disappointed with
those switches. This isn't the first time
they spat out mysterious behavior to basic
functions (MAC learning, populating the ARP
table). I ranted on the j-nsp mailinglist
about this earlier and apparently I'm not
the only one having these issues. Someone
from Hosteurope confirmed having similar
problems. A couple of cheap HP procurve
3400cl-24G switches using the preinstalled
firmware worked for about 3 years in the
same scenario and did not need to be re-
started even once. After replacing them
with the EX-series things started to, well,
behave irrationally.

I truly hope that Juniper is going to put
some extra effort into bugfixing their
EX-series. I really like the usability of
JUNOS and would look forward to replacing
more of our legacy switching equipment with
the EX-series. Being aware that switching
is a relatively new field in the portfolio
of Juniper, I acknowledge that stuff like
this is likely to happen to the early
adopters. Still, it leaves some sort of bad
aftertaste if the basics are prone to being
flaky.

Or it's just me again... For some reason
unbeknownst to me, I tend to stumble
across bugs more often than other people ;)

If the lurking devs or SE's from Juniper
could provide some debugging mechanisms so
that useful information could be gather-
ed for you folks, I am all ears (yes, the
hardware sports an active support contract).

Thanks for enduring my ramblings.

Best regards,

Mit freundlichen Gruessen,

i. A. Sven Juergensen

- --
Fachbereich Netze/Projekte

KielNET GmbH
Gesellschaft fuer Kommunikation
Preusserstr. 1-9, 24105 Kiel

Telefon : 0431 2219-053
Mobil   : 0170 403 5600
Telefax : 0431 2219-005
E-Mail  : s.juergensen at kielnet.de
Internet: http://www.kielnet.de

Geschaeftsfuehrer Eberhard Schmidt
HRB 4499 (Amtsgericht Kiel)

PGP at
http://pgp.kielnet.de/sjuergensen/

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)

iEYEARECAAYFAkmGrNMACgkQnEU7erAt4TIQUwCg9FdkIaxPd9PDDA6l6lSqfNZ7
B1QAnjea4uZWW2pA3O2znhy0Vvph3xOU
=sKyz
-----END PGP SIGNATURE-----