mozilla / source

DEPRECATED - Code for source.mozillaopennews.org/
https://source.opennews.org
BSD 3-Clause "New" or "Revised" License
36 stars 27 forks source link

Search trouble #216

Closed kissane closed 11 years ago

kissane commented 11 years ago

Slifty reports that our search is having some trouble—a search for "data" produces a Page Not Found error: http://source.mozillaopennews.org/en-US/search/?q=data

ryanpitts commented 11 years ago

Weird! Searches for other terms appear to work as expected:

http://source.mozillaopennews.org/en-US/search/?q=boyer http://source.mozillaopennews.org/en-US/search/?q=django

I'll see if I can figure out what's up here.

ryanpitts commented 11 years ago

Also strange, a search for the term "data" works fine locally as well as in our staging environment:

https://source-dev.mozillalabs.com/en-US/search/?q=data

I may need to see if someone can rustle up some logs for the live server for this.

kissane commented 11 years ago

Yeah, everything else seems completely normal. Super odd.

On Tue, Sep 10, 2013 at 4:16 PM, Ryan Pitts notifications@github.comwrote:

Weird! Searches for other terms appear to work as expected:

http://source.mozillaopennews.org/en-US/search/?q=boyer http://source.mozillaopennews.org/en-US/search/?q=django

I'll see if I can figure out what's up here.

— Reply to this email directly or view it on GitHubhttps://github.com/mozilla/source/issues/216#issuecomment-24190820 .

ryanpitts commented 11 years ago

I suspect it's going to end up being a character encoding issue in a recent article that includes the term "data" ...

kissane commented 11 years ago

That sounds like an obnoxious problem to find.

On Tue, Sep 10, 2013 at 4:25 PM, Ryan Pitts notifications@github.comwrote:

I suspect it's going to end up being a character encoding issue in a recent article that includes the term "data" ...

— Reply to this email directly or view it on GitHubhttps://github.com/mozilla/source/issues/216#issuecomment-24191490 .

ryanpitts commented 11 years ago

OK, so I tested locally with a bunch of weird characters in the summary field, etc., but I couldn't get anything to fail. Thought maybe it had to do with queries returning too many results (not sure how much memory is allocated to elasticsearch), so tried some other queries for things like "journalism." All worked. Until I tried "maps," which failed just like "data" did.

Hit the admin, looked at articles that would have those terms in them, trying to figure out what the anomaly was. Got really excited when I saw Bogotá in a recent summary, but turns out that wasn't it. Then I landed on an article that we had removed from the site, and I started wondering if maybe there was still a record of it in the elasticsearch index, and that might be causing an error somewhere.

Tried a few things but still couldn't replicate the error locally. For fun, ran the update_index command on production with a --remove flag to kill any record of objects that are gone, and ... HEY. Works again. So that must have been it.

Woo http://source.mozillaopennews.org/en-US/search/?q=data

kissane commented 11 years ago

Woohoo!

On Tue, Sep 10, 2013 at 6:17 PM, Ryan Pitts notifications@github.comwrote:

OK, so I tested locally with a bunch of weird characters in the summary field, etc., but I couldn't get anything to fail. Thought maybe it had to do with queries returning too many results (not sure how much memory is allocated to elasticsearch), so tried some other queries for things like "journalism." All worked. Until I tried "maps," which failed just like "data" did.

Hit the admin, looked at articles that would have those terms in them, trying to figure out what the anomaly was. Got really excited when I saw Bogotá in a recent summary, but turns out that wasn't it. Then I landed on an article that we had removed from the site, and I started wondering if maybe there was still a record of it in the elasticsearch index, and that might be causing an error somewhere.

Tried a few things but still couldn't replicate the error locally. For fun, ran the update_index command on production with a --remove flag to kill any record of objects that are gone, and ... HEY. Works again. So that must have been it.

Woo http://source.mozillaopennews.org/en-US/search/?q=data

— Reply to this email directly or view it on GitHubhttps://github.com/mozilla/source/issues/216#issuecomment-24199317 .

ryanpitts commented 11 years ago

Apologies for the long and winding description, but I like to leave Future Ryan notes about fixing things in case they happen again ;)

kissane commented 11 years ago

I actually really love the long explanations. They increase the chances that I'll actually remember/recognize something if a similar thing happens later/elsewhere.

On Tue, Sep 10, 2013 at 10:10 PM, Ryan Pitts notifications@github.comwrote:

Apologies for the long and winding description, but I like to leave Future Ryan notes about fixing things in case they happen again ;)

— Reply to this email directly or view it on GitHubhttps://github.com/mozilla/source/issues/216#issuecomment-24208618 .