Investigation - are we updating `analysis:last-modified-at` as we should?

[Pierlou]

The IRVE consolidation had us notice something weird for this resource :

• hydra crawled it as expected regularly, and I even forced a crawl (thinking it would fix the problem) so we have "created_at": "2024-11-19T13:17:33.360363+00:00", as a last check
• however the resource was last modified lately :

    "internal": {
        "last_modified_internal": "2024-11-19T13:17:30.963000+00:00"
    },
    "last_modified": "2024-11-19T13:17:30.963000+00:00",

• this is also shown in the headers of the resource's URL :

        "headers": {
            "last-modified": "Tue, 19 Nov 2024 13:17:26 GMT",
        },

• BUT in the extras of the resource we still have "analysis:last-modified-at": "2021-08-27T11:26:29+00:00", 🤔

[bolinocroustibat]

I could reproduce. I'm on it

[bolinocroustibat]

After inquiring, it seems the issue is related to the sending of analysis result to udata here, as I could verify the change_payload in send() method did include the updated "analysis:last-modified-at": "2021-08-27T11:26:29+00:00".

So it could be either:

• udata is responding with a 404. Since this commit, we decided no error would appear in Sentry in that case
• udata is not properly taking the payload into consideration as it should anymore?

[bolinocroustibat]

Update:
Would be that related to the many errors 502 Bad Gateway that Hydra gets from udata?
https://errors.data.gouv.fr/organizations/sentry/issues/2333/?project=21&referrer=release-issue-stream

It seems this error is very frequent and happens for the very first time on 2024/09/29 and started to rise from 2024/10/01, on unknown environment (could be prod, or elsewhere).
This period corresponds to

• when we spawned the new DB machine db-02 as backup
• we deployed hydra 2.0.1 on dev and demo (but not prod)