| Summary: | ASTERISK-23475: res_odbc crash | ||
| Reporter: | Tobias Gunkel (tgunkel) | Labels: | |
| Date Opened: | 2014-03-13 04:39:02 | Date Closed: | 2014-03-31 10:58:34 |
| Priority: | Major | Regression? | No |
| Status: | Closed/Complete | Components: | Resources/res_odbc |
| Versions: | 1.8.26.0 11.8.0 | Frequency of Occurrence | Occasional |
| Related Issues: | |||
| Environment: | Linux Kernel 3.12.9 x86_64, 2 x Intel Xeon CPU E5530 (16 cores overall) | Attachments: | ( 0) backtrace_20140310.txt ( 1) backtrace_20140311.txt |
| Description: | Hi,
we're running a rather heavy load setup with approximately 200-500 calls per minute on one server (average). Since we switched from MYSQL() to res_odbc we ran into sporadic Asterisk crashes (about 3 times a day). Those crashes are not necessarily bound to cpu load. It happens at 40% cpu-idle as well as 95% cpu-idle. The last output from "asterisk-rcgvv" after the crash is always: NOTICE[433]: res_odbc.c:1563 odbc_obj_connect: res_odbc: Connected to asterisk [asterisk] I've uploaded two backtraces: http://pastebin.com/4M9jrc3g http://pastebin.com/Te5z6kkb All phone numbers were substituted with 0123456789. Our /etc/asterisk/res_odbc.conf looks like this: [asterisk] enabled => yes dsn => asterisk username => asterisk password => *** pooling => yes limit => 100 pre-connect => yes sanitysql => set wait_timeout = 86400 idlecheck => 1800 share_connections => yes negative_connection_cache => 120 [asteriskIB] enabled => yes username => asterisk dsn => asteriskIB password => *** pooling => yes limit => 50 pre-connect => yes sanitysql => select 1 idlecheck => 1800 share_connections => yes negative_connection_cache => 120 Switching back to MYSQL() immediately fixes the problem. Any help in debugging this issue would be greatly appreciated! Best regards, Tobias | ||
| Comments: | By: Tobias Gunkel (tgunkel) 2014-03-13 04:46:44.990-0500 Uploaded backtraces directly. By: Matt Jordan (mjordan) 2014-03-13 06:26:28.375-0500 Which distribution of Linux are you using? What version of {{libodbc}} is installed? As this is a crash in {{libodbc}}, the bug is likely to be in that library. By: Tobias Gunkel (tgunkel) 2014-03-13 06:48:46.221-0500 Debian 7.4 and libmyodbc (5.1.10-2+deb7u1), libodbc1 (2.2.14p2-5), unixodbc (2.2.14p2-5) By: Tobias Gunkel (tgunkel) 2014-03-13 08:48:17.469-0500 Ok, we will be testing a newer libodbc with unixODBC-2.3.2 tomorrow. Thanks for the hint! -Tobi By: Walter Doekes (wdoekes) 2014-03-17 04:49:09.133-0500 This could be a threading issue where you have a non-thread safe myodbc, or one that is in non-thread-safe mode. Try using pooled connections in res_odbc.conf. (Or look at the myodbc options for Threading = N.) By: Tobias Gunkel (tgunkel) 2014-03-18 05:30:44.185-0500 Looks promising so far. No crashes with unixODBC-2.3.2 since 3 days. I'm going to respond again in a few days, when I have more convincing results. By: Tobias Gunkel (tgunkel) 2014-03-31 02:01:48.653-0500 I think this issue can be closed. No more crashes since upgrade to unixODBC-2.3.2. Thank you for your help! By: Walter Doekes (wdoekes) 2014-03-31 02:28:14.286-0500 Tobias: cool. Can you hit "Send back" so the moderators can close it? Thanks. By: Tobias Gunkel (tgunkel) 2014-03-31 02:30:25.423-0500 I think this issue can be closed. No more crashes since upgrade to unixODBC-2.3.2. Thank you for your help! | ||