Edouard,

Yes, it is fixed by Ajay in mainline. We are planning to back port the fix to 
at least 2.21

Sachin

On Jun 26, 2015, at 1:28 AM, Édouard Thuleau 
<[email protected]<mailto:[email protected]>> wrote:

In fact, when I tried to fix that bug on master branch, I found that already 
fixed by Ajay in that patch [1].

[1] https://review.opencontrail.org/#/c/11812/

Édouard.

On Fri, Jun 26, 2015 at 9:40 AM, Édouard Thuleau 
<[email protected]<mailto:[email protected]>> wrote:
Thanks Sachin.

Do you know if a fix already exists? Should I propose one?
I saw you made a patch [1] since 13th of June that seems concern the same type 
of error. I can not read the bug #1464901 since it's not public. Could publish 
it, please?

[1] https://review.opencontrail.org/#/c/11592
Édouard.

On Thu, Jun 25, 2015 at 6:50 PM, Sachin Bansal 
<[email protected]<mailto:[email protected]>> wrote:
Hi Edouard,

This could happen with following sequence of events:

1. Config node 1 gets a request to delete an object.
2. Config node 2 gets a request to update a field in the same object.
3. Config node 1 deletes the row in cassandra.
4. Config node 2 writes the row again in cassandra with only the field 
available in the request.

After this, the object is not in a consistent state in cassandra. We should 
catch the KeyError in cassandra read and return NoIdError.

Sachin


On Jun 25, 2015, at 8:53 AM, Édouard Thuleau 
<[email protected]<mailto:[email protected]>> wrote:

Hi Ajay,

We faced a schema transformer crash on our production (release 1.10) last 
Thursday due to internal error on the API process.

>From the python backtrace (schema [1], API server [2]), the schema sent an API 
>request to update a VMI and when the API server tried to build the response by 
>reading the VMI data with the VMI UUID key from cassandra db, it failed due to 
>the fq_name attribute missing and returned HTTP code 500 to schema.

Do you know how that's possible we had a resources in our database without an 
fq name?

[1] http://paste.openstack.org/show/321035/
[2] http://paste.openstack.org/show/321032/

Regards,
Édouard.




_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org

Reply via email to