Christian Schwarz
Christian Schwarz
Sorry for the delay in reviewing this PR. > This is based on the structure that the prometheus/exporter-toolkit web-config uses. Why not pull in the `prometheus/exporter-toolkit`s `TLSConfig` instead of rolling...
Hmm, I guess I'll have to bump a lot of dependencies first. I will do that in a separate PR, but to unblock you, just do whatever bumping you need...
For the record, WIP branch with dependency updates here: https://github.com/zrepl/zrepl/pull/819
FYI, #819 was merged a couple of days ago, so, minimum Go version is 1.22 now. Also, I restructured the repository and moved virtually all of the code into `internal/`....
Transitioning this back to draft state, please reopen and ping me once ready.
> I'm sure we can find a solution where the pageserver gets the feedback it needs regardless of hot_standby_feedback. Yeah, I think it's a one-line fix in pageserver. I can...
There was more discussion on Slack. https://neondb.slack.com/archives/C088Z5H1MPS/p1749685105258789?thread_ts=1739821964.736699&cid=C088Z5H1MPS I proposed the following solution: > I think the correct way forward is to > 1. remove the standby_horizon stuff and > 2....
Given above decision to replace `standby_horizon` with leases, we're not fixing the bug pointed out in this issue.
For posterity, there is more nuance to this bug. Further discussion happening in https://databricks.atlassian.net/browse/LKB-372
I do not understand how this is not served by `root_fs` ATM: ``` jobs: - ... root_fs=backup/vmhost1 connect: # ... to vmhost 1 - ... root_fs=backup/vmhost2 connect: # ... to...