Hi Anish,
Thank you for your reply. I'm having the following problem when compiling:
...
Running virtualenv with interpreter testroot/bin/python
New python executable in analytics_test/bin/python
Installing
Setuptools.............................................................................................done.
Installing
Pip....................................................................................................................................done.
/bin/bash -c "source
/home/agutierrez/workspace/Contrail/opencontrail2/build/debug/analytics_test/bin/activate; pip install --download-cache=/tmp/cache/agutierrez/systemless_test greenlet==0.4.1 gevent==0.13.8 eventlet==0.9.17 testtools==0.9.21 fixtures==0.3.12 requests>=1.1.0 lxml==2.3.3 geventhttpclient==1.0a prettytable==0.7.2 psutil==0.4.1 redis==2.7.1 xmltodict==0.2 thrift==0.8.0 bottle==0.11.6 mock==1.0.1 kafka-python==0.9.2 kazoo==1.3.1
stevedore"
scons: *** [build/debug/opserver/test/greenlet] Error 1
scons: building terminated because of errors.
Do you have any idea that what can be the problem or if there is any
log file that could provide more information about the error?
Best regards,
Alberto.
Quoting Anish Mehta <[email protected]>:
Hi Alberto,
Adding your own table is possible.
You need to add some entries to /controller/src/analytics/viz.sandesh
It does not need the collector to be rebuilt, but it does require
rebuilding contrail-analytics-api.
I'm going to assume the table name "ObjectAlberto" (instead of
"ObjectTableName")
I'm going to make this UVE accessible via the Analytics API as follows:
http://xx.xx.xx.xx:8081/analytics/uves/albertos
I'm going to make the ObjectLog visible via the Analytics Query API as
well (as you mentioned in your email)
Based on this API, from the command line of the analytics node you will be
able to do this:
contrail-logs --object-type alberto
1. Changes in controller/src/analytics/viz.sandesh
const string ALBERTO_TABLE = "ObjectAlberto"
const string ALBERTO_DISPLAY_NAME = "Alberto"
...
const map<string, string> UVE_MAP = {
...
"alberto" : ALBERTO_TABLE,
}
...
const map<string, objtable_info> _OBJECT_TABLES = {
ALBERTO_TABLE : {
'objtable_display_name' : ALBERTO_DISPLAY_NAME
'log_query_name' : "alberto"
}
}
2. Rebuild contrail-analytics-api
sb/controller/src$ scons -u opserver
3. Install and test on target
You should rebuild the contrail-analytics package and install it on your
target.
For this kind of change, there is an alternative installation method as
well, given below:
This is the generated file that will change when you rebuild:
sb/controller/src$ ls
../../build/debug/opserver/opserver/sandesh/viz/constants.py
You can copy it onto your analytics node.
It will be at a location such as:
/usr/lib/python2.7/dist-packages/opserver/sandesh/viz/constants.py
After overwriting the file (you should take a backup first, just in case),
restart contrail-analytics-api:
service contrail-analytics-api restart
4. Test the UVE Analytics API and Analytics Query API after starting your
application and connecting to the collector.
Regards,
Anish
On 4/23/15, 1:34 AM, "[email protected]" <[email protected]> wrote:
Hello Anish,
Thanks for your reply. What we are going to do is a little application
that sends it's own metrics to the collector so we can retrieve that
data using de Analytics API.
For example:
struct TestAgent {
1: string name(key="ObjectTableName")
2: optional i32 valueToBeSent1
}
uve sandesh TestTrage {
1: TestAgent data;
}
This kind of message should be stored in some way in Cassandra
databases and be available to the API users. I understand that
(key="ObjectTableName") makes reference to a new table with that name
that can be queried from the API. For me that is the concept of
"custom table", tables that are administrator/developer defined
different from base contrail tables.
It would be retrieved with a POST query to the API similar to this:
{
"end_time": "now",
"select_fields": [
"valueToBeSent1"
],
"sort_fields": [],
"start_time": "now-60m",
"table": "ObjectTableName",
"where": [
[
{
"name": "Source",
"op": 1,
"suffix": null,
"value": "GeneratorApplication",
"value2": null
}
]
]
}
Is this possible? If it is, do I have to do any step apart from send
the structure using the code generated by Sandesh compiler?
If it is not possible, is there any workaround? The only way that
comes to my mind is that we could query directly to message table,
filter the messages we don't want and process the XMLs but that
solution wouldn't be clean.
I hope I explained it correctly. Thank you for your assistance.
Best regards,
Alberto Gutiérrez.
Quoting Anish Mehta <[email protected]>:
The schemas for new Sandesh types do not need to be compiled or loaded
onto the Collector.
The collector gets XML-encoded messages from the generators and stores
them according to their sandesh types and the contained annotations.
Sandesh Types and how they work are explained here:
http://www.opencontrail.org/sandesh-a-sdn-analytics-interface/
I¹m not sure what you mean by ³custom tables².
Please send us a sample sandesh file and describe what you are trying to
do, and we can discuss further.
Regards,
Anish
On 4/22/15, 6:53 AM, "[email protected]" <[email protected]> wrote:
Dear all,
When creating a new sandesh structure, do I need to send somehow the
schema to the collectors so they will understand it or it will be
added automatically with the first message it recieves?
Thanks for your attention,
Alberto.
_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.or
g
_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org
