Yeah, I already have it disabled… was hoping maybe it might have been fixed since I’ve got a lot of Cisco stuff I watch.
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Tuesday, August 18, 2015 10:14 AM To: Observium Network Observation System Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
I guess this is just because cisco kit is /super/ slow at fdb-table for other reasons. We strongly recommend disabling it on these devices :)
Adam.
Sent with AquaMail for Android http://www.aqua-mail.com
On 18 August 2015 3:38:50 pm "Ron Marosko" ron@rjr-services.com wrote:
Yeah, minor brainfart on specifics… it was a rough thing and I can’t say I saw much change… refresh me on how to svn back down to the previous version and I’ll give you exact numbers for the module and on which svn release. I’m on the “stable” train.
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Tuesday, August 18, 2015 2:47 AM To: Observium Network Observation System Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
Cisco is a special case since it requires per-vlan context polling.
Also, your numbers are terribly confusing. Pre and post change, but you give numbers for with and without the module, not numbers for the module pre and post change.
Adam.
Sent with AquaMail for Android http://www.aqua-mail.com
On 18 August 2015 00:15:12 "Ron Marosko" ron@rjr-services.com wrote:
Just for some comparative numbers, both pre and post change… polling a Cisco 7609 without fdb-table takes 15-20 seconds. With fdb-table, it takes 130-140 seconds.
…Ron
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Monday, August 17, 2015 3:37 PM To: observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
These changes are now in both trunk and stable. I've disabled the table output until we can investigate the performance of large table outputs.
adam.
On 17/08/2015 20:59:08, Bill Fenner fenner@gmail.com wrote:
I still had another device that had worse fdb-table performance time than I would expect, and found a little bug in the code that updates fdb_status. If you ever end up with an entry that changes its fdb_status value (e.g., from "learned" to "invalid" before it gets aged out), this code causes the new value to be '' instead of the new value; then every subsequent run finds that fdb_status is wrong and attempts to rewrite it but rewrites it to ''. The diff:
http://www.fenron.com/~fenner/observium-update-fdb-status.diff
This brought this device's fdb-table down from 40s to 18s.
Bill
On Mon, Aug 17, 2015 at 3:41 PM, Adam Armstrong adama@memetic.org wrote:
Ahh. Thanks Bill! This would explain it.
I don't have any devices with large numbers of fdb entries, so I wouldn't have caught the abysmal performance.
I'll disable the table mode for this module.
Adam.
Sent with AquaMail for Android http://www.aqua-mail.com
On 17 August 2015 20:15:37 Bill Fenner fenner@gmail.com wrote:
I've been disabling the fdb-tables poller in my observium instances due to performance problems for a while. Adam poked me to look into this, so I root caused it to the code that displays the contents of the fdb to the terminal. You can apply
http://www.fenron.com/~fenner/observium-print-cli-table-timing.diff
to see this effect - it prints out the time taken to print the table if it takes more than 2 seconds. In my test case, it takes around 30 seconds to print a 2800-entry fdb (and the fdb-table module takes around 33 seconds).
So, it's nothing Arista-specific - it's some behavior of the table printer. The workaround is to comment out the calls to print_cli_table() in includes/polling/fdb-table.inc.php . Polling my sample Arista with about 2800 FDB entries, the module takes 33.4 seconds with the print_cli_table() calls and 6 seconds without.
Bill
On Mon, Aug 17, 2015 at 10:02 AM, Aaron Mayfield amayfield@artisaninfrastructure.com wrote:
Is there anyone else out there polling Arista switches? Has anyone else out there noticed any performance issues with polling the fdb-tables module recently?
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Aaron Mayfield Sent: Thursday, August 13, 2015 5:54 PM
To: Observium Network Observation System observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
OK, I picked one of the Arista switches at random and collected the poller performance stats both with and without fdb-table enabled. Attaching screenshots.
Thanks
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Thursday, August 13, 2015 4:48 PM To: observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
You also want the device performance for one of these arista devices, where you'll see which module is taking the time.
It's on the right hand side of the device navbar.
adam.
On 13/08/2015 22:46:05, Aaron Mayfield amayfield@artisaninfrastructure.com wrote:
Adam, you also asked for screenshots of the polling performance page. My apologies for taking so long to grab these.
This kind of confirms my suspicion that it is my Arista devices that are taking longer, they have the “swa” in the hostname.
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Wednesday, August 12, 2015 3:58 AM To: Observium Network Observation System observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
Hi Aaron,
Can you tell which device is taking a long time?
You can check the poller performance page from the "globe" menu, and the device performance tab (the "clock" icon on the right of the device navbar)
Screenshots of those might help :)
Thanks, Adam.
Sent with AquaMail for Android http://www.aqua-mail.com
On 12 August 2015 7:50:28 am Aaron Mayfield amayfield@artisaninfrastructure.com wrote:
Yeah this seems to be a performance issue of some type and the fdb table stuff seems like a side effect. I’m scratching my head as to why it worked fine before the upgrade. In fact, here is the observium.log output before/after the upgrade:
[2015/08/11 10:42:14 -0500] poller-wrapper.py(23384): /opt/observium/poller-wrapper.py: polled 45 devices in 131 seconds with 8 workers
[2015/08/11 10:47:12 -0500] poller-wrapper.py(3121): /opt/observium/poller-wrapper.py: polled 45 devices in 131 seconds with 8 workers
[2015/08/11 10:52:13 -0500] poller-wrapper.py(15078): /opt/observium/poller-wrapper.py: polled 45 devices in 132 seconds with 8 workers
[2015/08/11 10:57:13 -0500] poller-wrapper.py(27618): /opt/observium/poller-wrapper.py: polled 45 devices in 131 seconds with 8 workers
[2015/08/11 11:02:14 -0500] poller-wrapper.py(7205): /opt/observium/poller-wrapper.py: polled 45 devices in 132 seconds with 8 workers
[2015/08/11 11:07:14 -0500] poller-wrapper.py(19611): /opt/observium/poller-wrapper.py: polled 45 devices in 132 seconds with 8 workers
[2015/08/11 11:12:12 -0500] poller-wrapper.py(31781): /opt/observium/poller-wrapper.py: polled 45 devices in 131 seconds with 8 workers
[2015/08/11 11:17:15 -0500] poller-wrapper.py(11383): /opt/observium/poller-wrapper.py: polled 45 devices in 133 seconds with 8 workers
[2015/08/11 11:22:15 -0500] poller-wrapper.py(23688): /opt/observium/poller-wrapper.py: polled 45 devices in 133 seconds with 8 workers
[2015/08/11 11:27:14 -0500] poller-wrapper.py(3412): /opt/observium/poller-wrapper.py: polled 45 devices in 133 seconds with 8 workers
[2015/08/11 11:32:10 -0500] poller-wrapper.py(15327): /opt/observium/poller-wrapper.py: polled 45 devices in 128 seconds with 8 workers
[2015/08/11 11:37:14 -0500] poller-wrapper.py(27814): /opt/observium/poller-wrapper.py: polled 45 devices in 132 seconds with 8 workers
[2015/08/11 11:42:13 -0500] poller-wrapper.py(7491): /opt/observium/poller-wrapper.py: polled 45 devices in 132 seconds with 8 workers
[2015/08/11 11:47:13 -0500] poller-wrapper.py(19987): /opt/observium/poller-wrapper.py: polled 45 devices in 131 seconds with 8 workers
[2015/08/11 11:52:15 -0500] poller-wrapper.py(32100): /opt/observium/poller-wrapper.py: polled 45 devices in 133 seconds with 8 workers
[2015/08/11 11:57:14 -0500] poller-wrapper.py(11743): /opt/observium/poller-wrapper.py: polled 45 devices in 133 seconds with 8 workers
[2015/08/11 12:02:22 -0500] poller-wrapper.py(23906): /opt/observium/poller-wrapper.py: polled 45 devices in 140 seconds with 8 workers
[2015/08/11 12:06:52 -0500] poller-wrapper.py(4395): /opt/observium/poller-wrapper.py: polled 45 devices in 111 seconds with 8 workers
[2015/08/11 12:21:24 -0500] poller-wrapper.py(11770): /opt/observium/poller-wrapper.py: polled 45 devices in 683 seconds with 8 workers
[2015/08/11 12:38:09 -0500] poller-wrapper.py(17020): /opt/observium/poller-wrapper.py: polled 45 devices in 1388 seconds with 8 workers
[2015/08/11 12:48:30 -0500] poller-wrapper.py(26555): /opt/observium/poller-wrapper.py: polled 45 devices in 1708 seconds with 8 workers
[2015/08/11 13:06:30 -0500] poller-wrapper.py(2438): /opt/observium/poller-wrapper.py: polled 45 devices in 2487 seconds with 8 workers
[2015/08/11 13:12:30 -0500] poller-wrapper.py(9984): /opt/observium/poller-wrapper.py: polled 45 devices in 2548 seconds with 8 workers
[2015/08/11 13:31:56 -0500] poller-wrapper.py(19437): /opt/observium/poller-wrapper.py: polled 45 devices in 3414 seconds with 8 workers
[2015/08/11 13:40:50 -0500] poller-wrapper.py(25290): /opt/observium/poller-wrapper.py: polled 45 devices in 3647 seconds with 8 workers
[2015/08/11 13:55:34 -0500] poller-wrapper.py(956): /opt/observium/poller-wrapper.py: polled 45 devices in 4231 seconds with 8 workers
[2015/08/11 14:02:54 -0500] poller-wrapper.py(7354): /opt/observium/poller-wrapper.py: polled 45 devices in 4370 seconds with 8 workers
[2015/08/11 14:20:50 -0500] poller-wrapper.py(14288): /opt/observium/poller-wrapper.py: polled 45 devices in 5147 seconds with 8 workers
I did the upgrade and then the poller starts running poorly. Then I see the side effect of the of the duplicate entries with the fdb table because the poller processes are running so slowly they are stacking on top of each other.
So I have some kind of performance issue.
One strange thing, I only have 45 devices I’m polling, yet the port count under the Port menu says I have 44859 ports. That must be a miscalculation of some type or my database is messed up somewhere. Not sure if related or not.
Going to continue to try the poller manually and see if I can figure out where the slowdown is.
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Wednesday, August 12, 2015 1:33 AM To: observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
Man, why didn't I think of this?
This sounds like the problem. I guess something caused your poller processes to get confused, and they ended up running in parallel, not unthinkable when the same part of the poller process runs for so long, 74 seconds for the fdb-table module.
If you don't /really/ need this data, I'd recommend disabling it. It's one of the trade offs we have to make between performance and data.
If you still want the fdb data, you can force that module to be run using a less-often scheduled process in cron like ./poller.php -h all -m fdb-table
Note that this will break whatever graphs (fdb count?) that fdb-table produces, but you'll still have the data in the database.
adam.
On 12/08/2015 07:28:55, Tom Laermans tom.laermans@powersource.cx wrote:
If you're running multiple simultaneous pollers against the same device is not unthinkable they'll all be trying to insert the same data into the table...
Tom
On Aug 12, 2015 8:02 AM, Aaron Mayfield wrote:
Here is the requested output. I had to kill all the other poller processes running on the system to get it to run (they were hosing the CPU). Oddly enough, once I killed all the other processes, I didn’t any problems running it (back to being fast). Also don’t seem to be getting the errors in the db.log when running the poller ‘one-at-a-time’.
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Tuesday, August 11, 2015 11:52 PM To: observium@observium.org Subject: Re: [Observium] 'Duplicate entry' issues on vlans_fdb
Hi Aaron,
These seem to be gone from my db.log. Could you send me a ./poller.php -h 54 -m fdb-table -d ?
Thanks,
adam.
On 12/08/2015 05:01:53, Aaron Mayfield wrote:
Just today I updated to the latest and greatest (0.15.8.6882). I was several revisions behind and several database updates were done as a result. After the update, I noticed my poller.php processes started taking all the CPU, started getting gaps in the graphs, etc. I noticed thousands of these entries in db.log:
[2015/08/11 22:01:02 -0500] poller.php(21342): Failed dbQuery (#1062 - Duplicate entry '54-2083-6ef88537f91f' for key 'dev_vlan_mac'), Query: INSERT INTO `vlans_fdb` (`device_id`,`vlan_id`,`port_id`,`mac_address`,`fdb_status`) VALUES ('54','2083','220115','6ef88537f91f','learned') [2015/08/11 22:01:02 -0500] poller.php(21342): Failed dbQuery (#1062 - Duplicate entry '54-2084-005056a927c2' for key 'dev_vlan_mac'), Query: INSERT INTO `vlans_fdb` (`device_id`,`vlan_id`,`port_id`,`mac_address`,`fdb_status`) VALUES ('54','2084','220115','005056a927c2','learned') [2015/08/11 22:01:02 -0500] poller.php(21342): Failed dbQuery (#1062 - Duplicate entry '54-2084-a66aaf0bf4cc' for key 'dev_vlan_mac'), Query: INSERT INTO `vlans_fdb` (`device_id`,`vlan_id`,`port_id`,`mac_address`,`fdb_status`) VALUES ('54','2084','220115','a66aaf0bf4cc','learned') [2015/08/11 22:01:02 -0500] poller.php(21342): Failed dbQuery (#1062 - Duplicate entry '54-2085-228a3d193c66' for key 'dev_vlan_mac'), Query: INSERT INTO `vlans_fdb` (`device_id`,`vlan_id`,`port_id`,`mac_address`,`fdb_status`) VALUES ('54','2085','220115','228a3d193c66','learned') [2015/08/11 22:01:02 -0500] poller.php(21342): Failed dbQuery (#1062 - Duplicate entry '54-2086-0ee7c729643b' for key 'dev_vlan_mac'), Query: INSERT INTO `vlans_fdb` (`device_id`,`vlan_id`,`port_id`,`mac_address`,`fdb_status`) VALUES ('54','2086','220115','0ee7c729643b','learned')
If I run a poller process against a switch manually, everything seems to run fine with the exception of the fdb-table module, which is taking over 600 seconds to run.
Here is the schema of my vlans_fdb file:
mysql> show columns from vlans_fdb -> ; +-------------+-------------+------+-----+---------+-------+ | Field | Type | Null | Key | Default | Extra | +-------------+-------------+------+-----+---------+-------+ | device_id | int(11) | NO | PRI | NULL | | | vlan_id | int(11) | NO | PRI | NULL | | | port_id | int(11) | YES | MUL | NULL | | | mac_address | varchar(32) | NO | PRI | NULL | | | fdb_status | varchar(32) | NO | | NULL | | +-------------+-------------+------+-----+---------+-------+ 5 rows in set (0.00 sec)
mysql> mysql> show index from vlans_fdb -> ; +-----------+------------+--------------+--------------+-------------+-----------+-------------+---- ----+--------+------+------------+---------+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub art | Packed | Null | Index_type | Comment | +-----------+------------+--------------+--------------+-------------+-----------+-------------+---- ----+--------+------+------------+---------+ | vlans_fdb | 0 | dev_vlan_mac | 1 | device_id | A | 15 | ULL | NULL | | BTREE | | | vlans_fdb | 0 | dev_vlan_mac | 2 | vlan_id | A | 18348 | ULL | NULL | | BTREE | | | vlans_fdb | 0 | dev_vlan_mac | 3 | mac_address | A | 128440 | ULL | NULL | | BTREE | | | vlans_fdb | 1 | device_id | 1 | device_id | A | 78 | ULL | NULL | | BTREE | | | vlans_fdb | 1 | port_id | 1 | port_id | A | 431 | ULL | NULL | YES | BTREE | | +-----------+------------+--------------+--------------+-------------+-----------+-------------+---- ----+--------+------+------------+---------+ 5 rows in set (0.04 sec)
mysql>
Does my table structure look right? I see someone else on the list has had this same issue, but there is no indication that this should be a problem in the latest version.
What should I check? Thanks for any help.
Aaron Mayfield Cloud Expert Networking Specialist
12400 Hwy. 71 W. Suite 350-407 Austin, TX 78738 T. 512.600.4297 www.artisaninfrastructure.com Partner portal: https://portal.vpdc.us Partner support: support@artisaninfrastructure.com
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.) _______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.)
_______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.) _______________________________________________ observium mailing list observium@observium.org mailto:observium%40observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.)
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.)
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the system manager. Please note that any views or opinions present in this email are solely those of the author and do not necessarily represent those of the company. Finally the recipient should check this email and any attachment for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. (Proprietary & Confidential – Artisan Infrastructure, Inc. et all.)
_______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.org mailto:observium%40observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium MailScanner has detected a possible fraud attempt from "postman.memetic.org" claiming to be http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.org mailto:observium%40observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium MailScanner has detected a possible fraud attempt from "postman.memetic.org" claiming to be http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.org mailto:observium%40observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium MailScanner has detected a possible fraud attempt from "postman.memetic.org" claiming to be http://postman.memetic.org/cgi-bin/mailman/listinfo/observium