1267943 - Add ElasticSearch add-on to heroku

Status: RESOLVED FIXED

(Tree Management :: Treeherder: Infrastructure, defect)

Description

[James Graham [:jgraham]]

The plan for the next round of autoclassification improvements is to use ElasticSearch for looking up classifications. Therefore we need some ES add-on enabled in our Heroku instance.

Comment 1

[Ed Morley [:emorley]]

The available Elasticsearch addons are:

https://elements.heroku.com/addons/bonsai
https://elements.heroku.com/addons/foundelasticsearch
https://elements.heroku.com/addons/searchbox

We'll need to add it to all of {prototype instance, new stage, new prod}.

You should have permissions on all three, let me know if there are any problems :-)

Comment 2

[James Graham [:jgraham]]

On heroku:

mysql> select count(*), sum(length(job_guid)), sum(length(test)), sum(length(subtest)), sum(length(status)), sum(length(message)) from failure_line where action="test_result";
+----------+-----------------------+-------------------+----------------------+---------------------+----------------------+
| count(*) | sum(length(job_guid)) | sum(length(test)) | sum(length(subtest)) | sum(length(status)) | sum(length(message)) |
+----------+-----------------------+-------------------+----------------------+---------------------+----------------------+
|  1487816 |              59522018 |          97506130 |            109194076 |             6162712 |            523743064 |
+----------+-----------------------+-------------------+----------------------+---------------------+----------------------+

Adding in the other field data, it seems like there's about 750Mb of pure data that will go in ES. Of course there will be some overhead and need for growth. Based on that estimate, the number of documents, and the need for high availability, it looks like the options are:

https://elements.heroku.com/addons/foundelasticsearch - Bonsai 30 ($150/month)
https://elements.heroku.com/addons/foundelasticsearch - Dachs HA ($127/month)
https://elements.heroku.com/addons/searchbox - Advanced ($59/month)

These plans are ofc not equivalent in other dimensions.

Comment 3

[James Graham [:jgraham]]

I should also note that the 750Mb seems like a huge underestimate compared to the 33Gb table size, so I don't really trust it.

Comment 4

[James Graham [:jgraham]]

So I think I'm tempted to start with the foundelasticsearch smallest instance; although that doesn't have HA, it is offering a modern ES version whereas the others are very unclear about what they offer, and almost certainly has enough space for our needs. If it turns out that this is excessive it won't be a problem to go smaller later.

emorley: does that sound reasonable?

Comment 5

[James Graham [:jgraham]]

s/smallest/dachs/ if you will

Comment 6

[Ed Morley [:emorley]]

Yup, go for it :-)

Comment 7

[Ed Morley [:emorley]]

It's also worth noting that that addon is the one from the creators of Elasticsearch itself, whereas the others are from third parties. As such I think it's a definite winner.

Comment 8

[James Graham [:jgraham]]

OK, closing this out for now if we are happy with the choice. Seems to be plenty of space at least.

Comment 9

[Ed Morley [:emorley]]

Could you add it to the other two Heroku instances too please :-)

(Plus sync any config changes within the addon dashboard itself)

Comment 10

[James Graham [:jgraham]]

Added to the other instances now.

Comment 11

[Ed Morley [:emorley]]

Thank you :-)