lags reported by cmak was completely inconsistent

Good one, thanks.

Conditionally accepted on whether the inconsistency between _execute_with_retries and _execute_many_with_retries is worth it.

Build is green

Update: explicitely close response once processed

Build is green

Build was aborted

Fix test

In D6235#161236, @vlorentz wrote:

There's a bunch of typos in your commit/diff msg: "wich", "oef", "ony", "ALL offsets that needs to be", "stash" -> "squash"

---

this is necessary to ensure these messages are committed in kafka,
otherwise, since the (considered) empty partition is unsubscribed from,
it never gets committed in JournalClient.handle_messages() (since this
later only commit assigned partitions).

Why is this a problem?

Build is green

Update: also stream FTP responses.

rebase

I think this makes sense, but I'd appreciate @seirl's look on this as well. I made a few cosmetic comments regardless.

Seems to be a preexisting issue. Nevermind!

I'm tempted to postpone this issue until we resolve T2063...

A replacement disk will be sent by DELL. It should be delivered the 2021-09-13 if everything is ok.
The DSI is notified of the delivery

+1, I was also thinking recently to implement such exception filtering but you were faster.

There's a bunch of typos in your commit/diff msg: "wich", "oef", "ony", "ALL offsets that needs to be", "stash" -> "squash"

Can we keep the reset stuff outside the journalprocessor.py logic? It's already complex enough

I just tried it with an empty docker instance, and I'm getting this:

We have an icinga plugin implementing some of this: https://forge.softwareheritage.org/source/swh-icinga-plugins/browse/master/swh/icinga_plugins/vault.py

This is implemented by the git-bare cooker, but it's not publicly available for now.

This has been implemented and is now used by swh-web in production, closing this.

rebase

rebase

• fix the wrong box url
• Explain the move to nfs v4
• Explain the puppet directories hack

lgtm

final tests with the last version, everything looks good with almost the same performances a better ingestion rate in batch:
4 nodes before (batch only):

The blog post cannot be published on Tuesday 14 September as planned.
Pierre-Yves informs that: "It's not done yet, because we haven't discussed the memorandum yet".
I proposed to move forward on the blog post in order to be ready for publication once the whole thing has been published by NLNet (the memorandum).

Rebase

Build is green

add forgotten revision: Reduce the size of the progress bar

Rebase over production

Remove spurious print