Guys,
For those relying on Radius accounting generated by SER beware that the combination of RadAcctId-AcctSessionId is not at all unique. If you use MySQL for storage of radius accounting you will miss CDRs every now and then (because is a unique index) and the update queries will update calls from the past as a result, calls with huge duration are being generated.
The problem is generated by a combination of factors:
1. Some UAs reuse same call id among multiple calls (Grandstream systematically show this behaviour) 2. The Acct-Unique-Session-Id is not unique either, it repeats itself every now and then.
I would appreciate to see reactions for this problem from the developers.
Regards, Adrian
At 01:36 PM 6/2/2004, AG Projects support wrote:
Guys,
For those relying on Radius accounting generated by SER beware that the combination of RadAcctId-AcctSessionId is not at all unique. If you use MySQL for storage of radius accounting you will miss CDRs every now and then (because is a unique index) and the update queries will update calls from the past as a result, calls with huge duration are being generated.
The problem is generated by a combination of factors:
- Some UAs reuse same call id among multiple calls (Grandstream systematically show this behaviour)
Indeed, that's why serweb uses combination of callid with from and to-tags for accounting, something similar can be certainly used with Radius+mysql.
-jiri
We do that "as a dirty trick " but looking in my statistics those tags are repeating themselves at a far higher rate than Radius ids. It seems a matter of luck of having a unique combination.
If Acct-Unique-Session-Id would be unique it solve the problem at radius server level independent of broken UAs who are reusing Acct-Session-Id.
Do you know where is the Acct-Unique-Session-Id generated, is it radius client or SER ?
Adrian
On Jun 2, 2004, at 3:01 PM, Jiri Kuthan wrote:
At 01:36 PM 6/2/2004, AG Projects support wrote:
Guys,
For those relying on Radius accounting generated by SER beware that the combination of RadAcctId-AcctSessionId is not at all unique. If you use MySQL for storage of radius accounting you will miss CDRs every now and then (because is a unique index) and the update queries will update calls from the past as a result, calls with huge duration are being generated.
The problem is generated by a combination of factors:
- Some UAs reuse same call id among multiple calls (Grandstream
systematically show this behaviour)
Indeed, that's why serweb uses combination of callid with from and to-tags for accounting, something similar can be certainly used with Radius+mysql.
-jiri
Serdev mailing list serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
We do that "as a dirty trick " but looking in my statistics those tags are repeating themselves at a far higher rate than Radius ids. It seems a matter of luck of having a unique combination.
If Acct-Unique-Session-Id had been unique it would solve the problem at RADIUS server level independent of broken UAs who are reusing Acct-Session-Id (the SIP call id).
Do you know where is the Acct-Unique-Session-Id generated, is it radius client or SER ?
Adrian
On Jun 2, 2004, at 3:01 PM, Jiri Kuthan wrote:
At 01:36 PM 6/2/2004, AG Projects support wrote:
Guys,
For those relying on Radius accounting generated by SER beware that the combination of RadAcctId-AcctSessionId is not at all unique. If you use MySQL for storage of radius accounting you will miss CDRs every now and then (because is a unique index) and the update queries will update calls from the past as a result, calls with huge duration are being generated.
The problem is generated by a combination of factors:
- Some UAs reuse same call id among multiple calls (Grandstream
systematically show this behaviour)
Indeed, that's why serweb uses combination of callid with from and to-tags for accounting, something similar can be certainly used with Radius+mysql.
-jiri
Serdev mailing list serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
Hi all,
What software do you use for making sure that your sipserver(s) are online and working?
I'm on the lookout for something that I can use with Nagios (ie anything that I can run from a shellprompt).
I have tried sipsak but it's not working on solaris (some strange bind/socket operations that breaks the program)
I would like the same feature as sipsak has,
1. Register account 2. Call account 3. Check for invite
Doing an external call to a free number would also be nice, just wait for the 200 OK and then cancel the call.
Best regards,
Thomas Björklund
we usa nagios with sipsak on linux. I don't know how sipsak is doing on solaris. (sipsak author cc-ed). We register only. cancelling when a 200 comes back does not work, that's too late already.
-jiri
At 04:14 PM 6/2/2004, Tomas Björklund wrote:
Hi all,
What software do you use for making sure that your sipserver(s) are online and working?
I'm on the lookout for something that I can use with Nagios (ie anything that I can run from a shellprompt).
I have tried sipsak but it's not working on solaris (some strange bind/socket operations that breaks the program)
I would like the same feature as sipsak has,
- Register account
- Call account
- Check for invite
Doing an external call to a free number would also be nice, just wait for the 200 OK and then cancel the call.
Best regards,
Thomas Björklund
Serusers mailing list serusers@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serusers
-- Jiri Kuthan http://iptel.org/~jiri/
On Wed, 2 Jun 2004, Jiri Kuthan wrote:
we usa nagios with sipsak on linux. I don't know how sipsak is doing on solaris. (sipsak author cc-ed). We register only. cancelling when a 200 comes back does not work, that's too late already.
-jiri
Thanks Jiri,
I'll try to solve this with Nils outside of the mailinglist.
Best regards,
Thomas Björklund
Hello,
to keep you informed: the problem with Tomas server was that is does signale asymmetric (it sends responses from another port the the listening port).
I just committed the new release 0.8.9 of sipsak, which should fix this problem. Thomas: can you please verify and confirm if this new sipsak version works with your server or not?
Thanks and greetings Nils
On Thursday 03 June 2004 12:04, Tomas Björklund wrote:
On Wed, 2 Jun 2004, Jiri Kuthan wrote:
we usa nagios with sipsak on linux. I don't know how sipsak is doing on solaris. (sipsak author cc-ed). We register only. cancelling when a 200 comes back does not work, that's too late already.
-jiri
Thanks Jiri,
I'll try to solve this with Nils outside of the mailinglist.
Best regards,
Thomas Björklund
Serusers mailing list serusers@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serusers
On Sun, 6 Jun 2004, Nils Ohlmeier wrote:
I just committed the new release 0.8.9 of sipsak, which should fix this problem. Thomas: can you please verify and confirm if this new sipsak version works with your server or not?
Thanks and greetings Nils
Nils and the list,
I'm sorry to report that there still is some problems using it with the Cisco SIP Proxy server.
I'll take it offlist with ju Nils.
Best regards,
Thomas Björklund
At 10:03 AM 6/7/2004, Tomas Björklund wrote:
On Sun, 6 Jun 2004, Nils Ohlmeier wrote:
I just committed the new release 0.8.9 of sipsak, which should fix this problem. Thomas: can you please verify and confirm if this new sipsak version works with your server or not?
Thanks and greetings Nils
Nils and the list,
I'm sorry to report that there still is some problems using it with the Cisco SIP Proxy server.
may I suggest you to use SER -- I'm confident your problem will be easy to fix then.
Andy
Hello,
On Wednesday 02 June 2004 16:14, Tomas Björklund wrote:
What software do you use for making sure that your sipserver(s) are online and working?
I'm on the lookout for something that I can use with Nagios (ie anything that I can run from a shellprompt).
I have tried sipsak but it's not working on solaris (some strange bind/socket operations that breaks the program)
Please send me the exact error messages. I'll then try to fix it. But a few weeks ago it worked for me on Solaris. Anyway, sent me any input that i can try to fix it. I have no Solaris machine availbale.
I would like the same feature as sipsak has,
- Register account
- Call account
- Check for invite
Doing an external call to a free number would also be nice, just wait for the 200 OK and then cancel the call.
I'll not add any feature which establishes a dialog with another SIP UA, because sipsak has no SIP stack and the results would be probably horrible.
But with sipsak you can first register a user and then sipsak tries to call itself on this just before registered account. This should show if your normal INVITE processing is still working (Note: sipsak does not support record-routing yet).
Greetings Nils
At 03:37 PM 6/2/2004, Adrian Georgescu wrote:
We do that "as a dirty trick " but looking in my statistics those tags are repeating themselves at a far higher rate than Radius ids. It seems a matter of luck of having a unique combination.
I don't think that including tags is a dirty trick. Dialogs are idenfitied by triple callid-fromtag-totag, an INVITE may initiate multiple dialogs, and keeping the information complete is clean. This information should be relatively unique even if UAC fails to generate random values -- callee's From tag will be good then.
If Acct-Unique-Session-Id had been unique it would solve the problem at RADIUS server level independent of broken UAs who are reusing Acct-Session-Id (the SIP call id).
Do you know where is the Acct-Unique-Session-Id generated, is it radius client or SER ?
I'm not sure.
-jiri
I realize now that the triplet has tags from both user agents so it should be safe enough.
Thanks Jiri
Adrian
On Jun 3, 2004, at 1:11 AM, Jiri Kuthan wrote:
At 03:37 PM 6/2/2004, Adrian Georgescu wrote:
We do that "as a dirty trick " but looking in my statistics those tags are repeating themselves at a far higher rate than Radius ids. It seems a matter of luck of having a unique combination.
I don't think that including tags is a dirty trick. Dialogs are idenfitied by triple callid-fromtag-totag, an INVITE may initiate multiple dialogs, and keeping the information complete is clean. This information should be relatively unique even if UAC fails to generate random values -- callee's From tag will be good then.
If Acct-Unique-Session-Id had been unique it would solve the problem at RADIUS server level independent of broken UAs who are reusing Acct-Session-Id (the SIP call id).
Do you know where is the Acct-Unique-Session-Id generated, is it radius client or SER ?
I'm not sure.
-jiri
It is generated by the radiusclient library. I found the following using google:
http://lists.cistron.nl/pipermail/freeradius-devel/2002-September/003530.htm...
Maybe you could find some solution if you google more.
Jan.
On 02-06 15:37, Adrian Georgescu wrote:
We do that "as a dirty trick " but looking in my statistics those tags are repeating themselves at a far higher rate than Radius ids. It seems a matter of luck of having a unique combination.
If Acct-Unique-Session-Id had been unique it would solve the problem at RADIUS server level independent of broken UAs who are reusing Acct-Session-Id (the SIP call id).
Do you know where is the Acct-Unique-Session-Id generated, is it radius client or SER ?
Adrian
On Jun 2, 2004, at 3:01 PM, Jiri Kuthan wrote:
At 01:36 PM 6/2/2004, AG Projects support wrote:
Guys,
For those relying on Radius accounting generated by SER beware that the combination of RadAcctId-AcctSessionId is not at all unique. If you use MySQL for storage of radius accounting you will miss CDRs every now and then (because is a unique index) and the update queries will update calls from the past as a result, calls with huge duration are being generated.
The problem is generated by a combination of factors:
- Some UAs reuse same call id among multiple calls (Grandstream
systematically show this behaviour)
Indeed, that's why serweb uses combination of callid with from and to-tags for accounting, something similar can be certainly used with Radius+mysql.
-jiri
Serdev mailing list serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
Serusers mailing list serusers@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serusers