Skip to content

Mark builders lost after many transient failures and reconnect handling fixes#351

Merged
pulkomandy merged 4 commits intohaikuports:masterfrom
mmlr:transient_failures
Nov 27, 2025
Merged

Mark builders lost after many transient failures and reconnect handling fixes#351
pulkomandy merged 4 commits intohaikuports:masterfrom
mmlr:transient_failures

Conversation

@mmlr
Copy link
Copy Markdown
Member

@mmlr mmlr commented Nov 26, 2025

After 10 consecutive transient failures, mark the builder lost. This prevents errors during build setup to repeat indefinitely. Also fix the original issue that triggered the change above, a symptom seen there and a related reconnect handling fix.

mmlr added 4 commits November 26, 2025 01:51
If the state remains AVAILABLE, no reconnect will be attempted in the
next run, leading to the reuse of a closed connection.
This mostly reverts ca6c9a3. The change
tried to "reconnect" SFTP on each command. As it reused the existing SSH
connection, this merely spawned a new internal SFTP client on the remote
side and wouldn't actually reestablish any connections. The new code
lead to only partial handling of exceptions, producing somewhat cryptic
errors instead of showing the underlying issue.

The logging changes from that commit are left intact.

Also clean up two unused methods.
The RECONNECT state does not need further handling, as the reconnection
happens automatically. Having the builder removed from the active
builder list would spuriously lead to the "all builders lost" state even
when reconnects should have happened.
@waddlesplash
Copy link
Copy Markdown
Member

Looks good to me.

@pulkomandy pulkomandy merged commit 22c3e93 into haikuports:master Nov 27, 2025
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

3 participants