IBX-3863: Parametrized Solr HTTP Client timeout and max retries

[alongosz]

Question	Answer
JIRA issue	IBX-3863
Type	improvement
Target Ibexa version	v3.3+
BC breaks	no

When switching to Symfony HTTP Client to handle Solr queries in #235. I didn't introduce any possibility to configure Client's timeout nor number of max retries. There are some edge cases which show that this parameter might work-around performance issues.

Documentation

Solr Ibexa Solr Bundle uses Symfony HTTP Client to fetch and update Solr index. It's possible to configure timeout and number of max retries for that client using Solr Bundle's Semantic Configuration:

v3.3:

ez_search_engine_solr:
    http_client:
        timeout: 30
        max_retries: 5

v4.x:

ibexa_solr:
    http_client:
        timeout: 30
        max_retries: 5

QA

You can test timeout change by mocking Solr service

$ nc -vv -l 8983

and then trying to perform e.g. reindexing

php ./bin/console ibexa:reindex --processes 1 -n

See that modifying both timeout and max_retries affect the wait time and the amount of tries before failure.

Open questions

Given this rather should be avoided, preferring proper performance fixes, should we:

• Document those parameters?
• Provide Semantic configuration instead of parameters (only if we decide to document)

Checklist:

• Provided PR description.
• Tested the solution manually.
• Provided automated test coverage.
• Checked that target branch is set correctly.
• Asked for a review

[Steveb-p]

Semantic bundle configuration is almost always preferred and recommended by Symfony.

[Steveb-p]

What you can additionally do is to pass additional options to ibexa.solr.http_client factory.

What I mean specifically, we can pass extra.curl.CURLOPT_CONNECTTIMEOUT (extra contains client-specific configurations) so that if we are having trouble establishing a connection, then we're dropping the attempt earlier (like after 2 seconds maybe?).

This will help us differentiate between service being overloaded / responding longer vs. service being completely unavailable.

See https://symfony.com/doc/current/http_client.html#configuring-curlhttpclient-options

[alongosz]

What you can additionally do is to pass additional options to ibexa.solr.http_client factory.

What I mean specifically, we can pass extra.curl.CURLOPT_CONNECTTIMEOUT (extra contains client-specific configurations) so that if we are having trouble establishing a connection, then we're dropping the attempt earlier (like after 2 seconds maybe?).

This will help us differentiate between service being overloaded / responding longer vs. service being completely unavailable.

See https://symfony.com/doc/current/http_client.html#configuring-curlhttpclient-options

@Steveb-p I got the error when trying to set that option

Cannot set "CURLOPT_CONNECTTIMEOUT" with "extra.curl"

Moreover I think it will be the opposite of what we're trying to achieve. We need to extend connecting time in case of slow solr instance, but we want to avoid retries when instance is not running at all.

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
0.0% Duplication

[tomaszszopinski]

QA approved on IbexaDXP 3.3 commerce.