Open jindrichmynarz opened 9 years ago
@jindrichmynarz do you have log from UV (master)?
Do you mean /var/log/unifiedviews/backend/backend.log
?
master.log (in frontend), I am not sure where it is located.
It is in /var/log/unifiedviews/frontend/master.log
, but I cannot find anything related to this issue there. See here: https://gist.github.com/jindrichmynarz/effdbfb238fce38a3b2c.
@jindrichmynarz most probably the URL of UV (master) API is wrongly set.
In /etc/odn-simple/odn-ckan-ic/production.ini there should be something like this:
odn.uv.api.url = http://127.0.0.1:8080/master/api/1
odn.uv.timeout = 5
can you check if the url is correct (especially port)? Check if there is http or https.
try: curl http://127.0.0.1:8080/master/api/1
<- url from cfg file
should give proper UV api error (JSON):
{
"errorMessage":"human readable error, something like: unexpected error occured",
"technicalMessage":"HTTP 404 Not Found"
}
I have http://127.0.0.1:28080/master/api/1
(note the different port) set as odn.uv.api.url
in /etc/odn-simple/odn-ckan-ic/production.ini
. Requesting this URL returns HTTP/1.1 404 Not Found.
if the response isn't JSON, the UV (master) API isn't running / reachable @Jan-Marcek can you help with this?
If the UV API is not running, then it is weird that I can still associate pipelines to datasets in CKAN.
To associate pipeline you need the API function that threw time out error in the above log.
Oh you mentioned you have 86 pipelines ... maybe just the API call, that retrieves the pipelines, last longer than 5 sec?
Did you try increasing the timeout (odn.uv.timeout = 5
)?
My point is that if the API is not running, then it cannot throw timeout errors.
I think it is highly likely that the API call just takes longer than the timeout, as you say. For example, now I have lower load on UV and the requests no longer timeout.
@jindrichmynarz so whats the problem, increase the timeout in CKAN config file? @skrchnavy why is this assigned to me, this is clearly UV performance issue
I think the problem is that this operation takes UV longer than it should. As you say, it is a UV performance issue.
Reported in UnifiedViews/Core#531
When I try to associate a pipeline to a dataset in the internal catalogue, I go to "Pipelines" tab, click "Add pipeline" and select to add an existing pipeline "Couldn't get pipelines, probably UnifiedViews is not responding." error message is shown. UnifiedViews is running and responding, but it is slow and the query retrieve pipelines times out as can be seen in
/var/log/apache2/odn-ckan-ic.error.log
:I have 86 pipelines in my instance of UnifiedViews and the instance runs on a beefy VM (64 GB RAM), so retrieving these pipelines should be swift. Ideally, this should be fixed in UnifiedViews. Alternatively, timeout could be extended in the CKAN plugin that calls the API of UnifiedViews.
Context: I'm using ODN 1.1.0.