Asterisk 13.1-cert2 crahsing with SEGFAULT on outgoing calls

Asterisk Asterisk Support
1

Hi everyone!

First of all. We’re trying somenting that’s not fully supported, so if you don’t bother to deal with such environments, please do not waste time on my posting here :wink:
The reason for this post is to figure out, how we should proceed, so I would delighted I you could provide me with a few hints about how to progress with our issue here.

We are evaluating an Installation with a Digium TE820 Card being handed off to a virtual machine by VMware6. (Yes, it is the customer’s idea) The idea why theyr’re doing this is to get more hardware-independence and availability. The host is running just one VM (The Asterisk I am talking about) and the PCIX Card is exclusively passed to that VM.
I am not sure whether the problem we have is related to virtualization and therefore I’d like to know your highly estimated opinions. (My questions are at the end of this post – maybe I am just not seeing some obvious incompatibility)

Our Problem is that on outgoing calls Asterisk crashes with a SEGFAULT either during the call or immediately afterwards. Incoming calls work flawlessly. During normal operations the calls would be submitted by IAX (form a backend Asterisk), but during testing we have also tried directly registered SIP phones… I do not want to withhold the dmesg glory :wink::

[16760.709494] asterisk[2368]: segfault at 144 ip 00000000005199ec sp 00007f20899f1360 error 4 in asterisk[400000+29a000]

A few words about the environment:
Host:
VMware6
Guest (=Asterisk):
Debian Linux 3.19.8-amd64+ with a self-compiled 1000Hz kernel, original VMware Tools installed

Dahdi:
2.10.2+2.10.2

LibPri:
1.4.15

Asterisk
certified-asterisk-13.1-cert2
We’re using chan_sip, but have also compiled pjsip for future developmens

Pjproject
pjproject-2.4.5

PRI-Card:
Wildcard TE820 (5th Gen)

PRI: EuroISDN

My questions would be
• Do you think this venture is BS alltogether? I could go on for few opages about the customer’s reasoning, buit I guess their Basic iodea is well thought-through, so they want to konow if our Installation could be run in such a Setup. (The both frontend Asterisk-machines are handling approx 10.000 calls a day each)
• From all the versions above, could it be that I have without seeing it chosen some versions that are old or inherently incompatible?
• I have searched for some kind cumulative patches that could be available, as I have found some bug reports that describe similar issues. Unfortunately I haven’t found a suitable patch. The file certified-asterisk-13.1-cert2-patch.gz in the download directory just seems to update an existing cert-13.1 to 13.2.
Are there any ‘cumulative patches’ If yes, where can I find them?
• If I submitted the whole issue as a bug, would I p**s off the guys maintaining the code as my environment uses a virtualized card?
• I haven’t yet signed up to Asterisk Issue tracker. Is there a detailed description of which files (Coredump, strace-output) will have to be provided for submitting the bug and how to create them? As I am mainly a networking guy using Linux, my knowledge on debugging issues is on the weak side :wink:

Thanks in advance for your time
Greetings from Austria,
George

2

Information on getting the information to debug a crash can be found at wiki.asterisk.org/wiki/display/ … +Backtrace

I don’t understand why you would expect there to be cumulative patches from every version to every other, or even from every version to the latest. If you really want to jump a long way, it is easier to download the complete source.

Whilst it shouldn’t cause crashes, running hardware with 20ms latency requirements on a virtual machine doesn’t seem to be a good idea. Also running a 1kHz kernel on such a machine seems an overkill, as, without a lot of optimisation of the hosting environment, and probably even with it, timing jitter is likely to be a lot more than 1ms.

The main problem in getting a bug progressed, assuming you can reproduce it on the latest version, would be that no-one is likely to have a suitable test environment, so, unless the cause is obvious from the core dump and a quick analysis of the code, it may be difficult to reproduce.

3

Hi thanks for your reply!

As I said before, it’s a customer idea - your posting has helped me a lot in the sense that I have now some additional arguments from a third party (‘moves like spencer’ :wink: about the virtualization of such a system in general…

Regarding the pathces I have misunderstood the basic idea of how the Patches are handled- sorry. I thought there are cumulative patches that fix some things within a given Version that are publsihed on a regular basis. I did not intend to compile me up from some old version, just tried to stay within the certified version. :wink:

I will now try out the latest Version of Asterisk and check whether the problem persists!

Best regards,
George

4

Hi everyone!

FYI (in case some else is tasked of to do such a thing)
I updated to 13.6.0, everything runs like a charm. We processed 2700 calls on this system in one day - no customer complaints… So everything seems to be running smoothly