Skip to content

Conversation

@jthomson04
Copy link
Contributor

ETCD seems to be very sensitive to network and memory pressure. During the network and memory-intensive model loading process (especially when loading from ETCD), pings from the ETCD server can take several seconds to be delivered. To get around this, we lazily initialize ETCD, and only create our lease after the model loading has completed.

Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
@jthomson04 jthomson04 requested a review from a team as a code owner January 27, 2026 19:40
@github-actions github-actions bot added the feat label Jan 27, 2026
@grahamking
Copy link
Contributor

I don't think we should do this. Seem my comment here: #5690 (comment)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants