That is interesting, but I'm pretty sure that we must be experiencing different 
crashes with different underlying causes.  It is pretty clear that ours are 
related to external HSS/RADIUS.

In fact, there appear to be 2 separate bugs:

1. When a RADIUS request times out (or possibly when a number of RADIUS 
requests are made at roughly the same time and they all time out...the rate of 
requests seems to be a factor), the process responsible for external HSS 
(aaa_iwk) is at risk of segfaulting.  This shouldn't happen, but is also 
compounded by the fact that the EPC does not try to restart the process 
gracefully, but instead initiates a 5 minute countdown, after which it reboots 
itself.

2. When a RADIUS request times out, and the EPC marks that RADIUS server as 
being unavailable/down, the EPC will also decide (apropos of nothing) that the 
eNB that the request was generated from is also down (even though it's not) and 
will reset its SCTP/S1 link to it, kicking all UEs off in the process.

The running theory we have developed at this point is that often what happens 
is that #2 occurs first, which snowballs into a huge number of UE 
detach/reattaches and combines with #1 + a slow-to-respond RADIUS server to 
ultimately cause the crash and reboot (e.g., the EPC transmits an interim 
accounting packet to the RADIUS server for a given UE, gets no response, 
decides to flag the eNB as down, which kicks all UEs attached to that eNB 
off-line, and then when the eNB re-establishes S1 with the EPC, there is a 
sudden flood of UE attachment requests that all come in at the same time, which 
also are not responded to by the RADIUS server fast enough, which ultimately 
leads to the aaa_iwk process dying).

The next few days will hopefully confirm this, but we think we have worked 
around the issue that was causing the RADIUS server to sometimes be slow to 
respond, so hopefully things will stabilize for us now.  But these two bugs 
still need to get fixed, especially since one is a crasher (though they are 
both service-disruptive).

-- Nathan

From: [email protected] [mailto:[email protected]] On Behalf Of 
Gabriel Pike
Sent: Thursday, January 05, 2017 10:27 AM
To: [email protected]
Subject: Re: [Telrad] BreezeWAY EPC spontaneous reboots?

I have experienced 2 of these crash/reboots and I am using built in iHSS. Once 
the EPC rebooted and the other time it completely locked up and needed power 
cycled. I have tickets open about this and thus far have not found a cause. I 
may have to do the same thing you did and plug in a laptop via serial cable and 
hope to capture what is happening.

Regards,

Gabriel Pike
Network Engineering
MTCNA
DMCI Broadband, LLC<http://dmcibb.net/>
[email protected]<mailto:[email protected]>

877.936.2422
Ext. 103


From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]] On Behalf Of Matthew Carpenter
Sent: Tuesday, January 03, 2017 9:24 AM
To: Telrad List
Subject: Re: [Telrad] BreezeWAY EPC spontaneous reboots?

Yes, we are only using the built in iHSS of the EPC.

Matt Carpenter


On Tue, Jan 3, 2017 at 6:34 AM, Nathan Anderson 
<[email protected]<mailto:[email protected]>> wrote:
Okay, after seeing from the console that the AAA process is the one dying and 
that the BreezeWAY waits a whole 5 minutes after its death before it reboots 
itself, I was able to pinpoint the place in the tlsyslog where the failure is 
being referenced.  It appears that for whatever reason (which I will work on 
hunting down), the BreezeWAY is sometimes failing to get a response from our 
RADIUS server.  And that sometimes when this happens, the AAA process 
apparently segfaults.

40087:2016-12-30,22:51:37.352929:ERROR:0:06.06.00729:AAA:1241:accounting 
request timeout max retransmissions reached [MSID = 0010010000xxxxx]
40088:2016-12-30,22:51:39.352964:ERROR:0:06.06.00729:AAA:1241:accounting 
request timeout max retransmissions reached [MSID = 0010010000xxxxx]
40089:2016-12-30,22:51:39.353026:ERROR:0:06.06.00729:AAA:1241:No response from 
AAA server, after interim retransmission, [MSID=0010010000xxxxx], srvc_grp[0], 
AAA_IP[xx.xx.xx.xx]
40090:2016-12-30,22:51:39.353075:ERROR:0:06.06.00729:AAA:1241:Accounting 
Interim timeout - send CLR to S6A, UE [0010010000xxxxx]
40100:2016-12-30,22:51:39.361255:NOTICE:0:06.06.00729:PGWC:1289:Counters 
indication sent to AAA IWK
40102:2016-12-30,22:51:40.215186:ERROR:0:06.06.00729:AAA:1241:accounting 
request timeout max retransmissions reached [MSID = 0010010000xxxxx]
40109:2016-12-30,22:51:40.294878:ERROR:0:06.06.00729:AAA:1241:UE 
[0010010000xxxxx] does not exist!
40118:2016-12-30,22:51:41.384408:ERROR:0:06.06.00729:AAA:1241:Auth T2 failure 
[MSID = 0010010000xxxxx] AAA server address [xx.xx.xx.xx]
40119:2016-12-30,22:51:41.882861:ERROR:0:06.06.00729:AAA:1241:Auth T2 failure 
[MSID = 0010010000xxxxx] AAA server address [xx.xx.xx.xx]
40120:2016-12-30,22:51:42.215202:ERROR:0:06.06.00729:AAA:1241:accounting 
request timeout max retransmissions reached [MSID = 0010010000xxxxx]
40121:2016-12-30,22:51:42.215252:ERROR:0:06.06.00729:AAA:1241:No response from 
AAA server, after interim retransmission, [MSID=0010010000xxxxx], srvc_grp[0], 
AAA_IP[xx.xx.xx.xx]
40122:2016-12-30,22:51:42.215301:ERROR:0:06.06.00729:AAA:1241:accounting 
request timeout max retransmissions reached [MSID = 0010010000xxxxx]
40123:2016-12-30,22:51:42.215448:ERROR:0:06.06.00729:AAA:1241:Accounting 
Interim timeout - send CLR to S6A, UE [0010010000xxxxx]

40135:2016-12-30,22:51:42.240884:NOTICE:0:06.06.00729:SIGNAL:1241:tl_signal.c:signal_handler(65):Traceback
 data saved: process:aaa_iwk.out, pid:1241, sig:Segmentation fault

2016-12-30,22:51:42:ERROR: COREDUMP Generated: 
/mnt/bigstore/coredumps/core_1483138302_06.06.00729_aaa_iwk.out.1241.gz

40142:2016-12-30,22:51:44.881842:ERROR:0:06.06.00729:FORKER:907:forker_api.cpp:StateHandler(855):Subsystem
 AAA SERVICE (id:9) process dead

Hopefully the generated core dumps can aid Telrad engineers in debugging the 
issue.

If most everybody is either using the iHSS, or they are using external HSS but 
not experiencing issues with their RADIUS server (or the network between the 
EPC and the RADIUS server), then that would explain why we are getting hit with 
this bug and other are not.

-- Nathan

From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]<mailto:[email protected]>] On Behalf Of 
Nathan Anderson
Sent: Tuesday, January 03, 2017 4:10 AM

To: Telrad List
Subject: Re: [Telrad] BreezeWAY EPC spontaneous reboots?

We have had further crashes/reboots.  Yesterday it crashed twice within 10 
minutes, then stabilized and hasn't crashed since.

This time, though, I had a serial cable hooked up to the system, and this got 
logged to the console both times:

  + AAA SERVICE................................................ [DEAD]
  - PGWC SERVICE............................................... [STOP]
  - SGWC SERVICE............................................... [STOP]
  - S6A SERVICE................................................ [STOP]
  - MME SERVICE................................................ [STOP]
  - UPGRADE INTERFACE.......................................... [STOP]
WARNING : recevied Error Message with reason=21 and cause =207 from 
Application/OAM-CL
WARNING : recevied Error Message with reason=21 and cause =157 from 
Application/OAM-CL
WARNING : recevied Error Message with reason=21 and cause =107 from 
Application/OAM-CL
WARNING : recevied Error Message with reason=21 and cause =107 from 
Application/OAM-CL
  - CONFIGURATION AGENT........................................ [STOP]
  - BASE CONFD................................................. [STOP]
  - CONFD PHASE-0.............................................. [STOP]
  - FSTS....................................................... [STOP]
  - HWIF....................................................... [STOP]

#################################################################################
                       System failure condition detected!
                     POWER restart scheduled in 300 second(s)
#################################################################################
Forker timeout has expired. Reset the board...
Requesting Power On system reset...

So I guess the AAA process is failing for some reason.  We are using external 
HSS (RADIUS), so I presume it has something to do with that.

Guess I'll dig through our FreeRADIUS logs and then open a ticket...

-- Nathan

From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]] On Behalf Of Nathan Anderson
Sent: Saturday, December 31, 2016 2:09 PM
To: Telrad List
Subject: Re: [Telrad] BreezeWAY EPC spontaneous reboots?

If a full reset is what it takes to fix, I am familiar enough with the 
procedure that I can do it myself.

To me, it sounds like confd is bombing out, and other processes (mme, pgwc) are 
trying to talk to it, unsuccessfully (since it isn't running).  A few minutes 
later, I presume the 'forker' process sees the problem and issues a system 
reboot.

Yesterday was the first and only time it has done this, and it happened 4 
times.  We have not had a recurrence in 24 hours.

-- Nathan

From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]] On Behalf Of Matthew Carpenter
Sent: Saturday, December 31, 2016 8:00 AM
To: Telrad List
Subject: Re: [Telrad] BreezeWAY EPC spontaneous reboots?

​Have not had any issues like this with our 2 EPCs.

I would contact support (Nick) and see about doing a full factory default on 
the EPC and set it back up from scratch.
We had some odd issues with an eNB and that was the solution, at least for an 
eNB.

The error messages sounds more like confd is trying to start a process with 
parameters and its not working.

Matt Carpenter

​

On Sat, Dec 31, 2016 at 12:34 AM, Nathan Anderson 
<[email protected]<mailto:[email protected]>> wrote:
Okay, I take it back.  I found a clue in the 'tlsyslog' after all.  Before it 
reboots, I see several of these logged in there, interwoven with otherwise 
normally-expected messages:

MME:1229:MME - Failed to start the session with confd for operational params

(...or...)

PGWC:1287:Failed to start the session with confd for operational params

So it sounds like confd is dying for some reason, and then the watchdog kicks 
the box a few minutes later.

So I guess the question is, why is confd dying.

-- Nathan

-----Original Message-----
From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]<mailto:[email protected]>] On Behalf Of 
Nathan Anderson
Sent: Friday, December 30, 2016 10:27 PM
To: [email protected]<mailto:[email protected]>
Subject: [Telrad] BreezeWAY EPC spontaneous reboots?

So, we had a new one today.  One of our EPCs rebooted itself 4 times within the 
span of 90 minutes.

Yes, latest public code level (6.6 729).

Are we the only ones who have seen THIS happen??

I didn't observe this particular detail myself, but others whose eyeballs were 
trained on the physical BreezeWAY box at the time say that the alarm light went 
red when it stopped responding, sat like that for a few minutes, and then the 
box finally rebooted (presumably some sort of watchdog process).

Is there anything that I can look for to explain the reboots?  'show 
notification stream alarms' only show the 'device-is-up-and-running' event with 
nothing suspicious-looking showing up before that (or at least that managed to 
get committed to NVRAM before the reboot occurred).  Similarly, the tlsyslog 
file just abruptly ends with "-- SYSTEM STARTED --" and "NOTICE:Last restart 
type is POWERUP" with nothing suspicious-looking getting logged right before 
that.

I now have a serial cable hooked up to it in case it happens again and in case 
whatever is causing the crash logs something to the console that isn't getting 
written to disk for some reason.  But of course it hasn't happened again (it's 
been 6 hours since the last event).

We haven't made a change to the config on this thing since the 6.6 upgrade was 
installed last August or whenever that was.  So, whyyyyyyyy after months of 
stability is this happeningggggggg.

...actually, I take that back.  We increased the uplink AMBR value to help 
troubleshoot the eNB capacity issues we have been seeing.  But that's all.

Ugh.  If it isn't one thing...

--
Nathan Anderson
First Step Internet, LLC
[email protected]<mailto:[email protected]>

_______________________________________________
Telrad mailing list
[email protected]<mailto:[email protected]>
http://lists.wispa.org/mailman/listinfo/telrad

_______________________________________________
Telrad mailing list
[email protected]<mailto:[email protected]>
http://lists.wispa.org/mailman/listinfo/telrad



--
Matthew Carpenter
806-316-5071<tel:(806)%20316-5071> office
806-236-9558<tel:(806)%20236-9558> cell

[https://docs.google.com/uc?export=download&id=0BxDRq5UV7HPOaEM4LXVaVnk5cWM&revid=0BxDRq5UV7HPOTDdiVjM0TXRIc3ZzMXVUVDdDVjBiaFU0bHJNPQ]

_______________________________________________
Telrad mailing list
[email protected]<mailto:[email protected]>
http://lists.wispa.org/mailman/listinfo/telrad



--
Matthew Carpenter
806-316-5071 office
806-236-9558 cell

[https://docs.google.com/uc?export=download&id=0BxDRq5UV7HPOaEM4LXVaVnk5cWM&revid=0BxDRq5UV7HPOTDdiVjM0TXRIc3ZzMXVUVDdDVjBiaFU0bHJNPQ]
_______________________________________________
Telrad mailing list
[email protected]
http://lists.wispa.org/mailman/listinfo/telrad

Reply via email to