3 ms·
This is what I've done in the past, not sure why we need to "hack TCP" to get zero downtime. As an additional thing I do rolling deploys where I just create an
by nulltype 11y ago
This is what I've done in the past, not sure why we need to "hack TCP" to get zero downtime.
As an additional thing I do rolling deploys where I just create an entirely new VM and add it before removing the old one. This just means that I don't have to recycle a node that might potentially have some state on it.
- tangled 11y agoThe article is about restarting HAProxy without any downtime. HAProxy restarts are needed when adding new service instances or adjusting configuration options. This is a different and much harder problem than gracefully restarting load balanced service instances.
- jolynch 11y agoAs other posters have mentioned the solution in this link is actually suboptimal. I showed how the iptables solution leads to 1-3s latency for connections that establish during the restart but you can avoid it by getting a little more hardcore. See http://engineeringblog.yelp.com/2015/04/true-zero-downtime-haproxy-reloads.html http://engineeringblog.yelp.com/2015/04/true-zero-downtime-h... or http://inside.unbounce.com/product-dev/haproxy-reloads/ http://inside.unbounce.com/product-dev/haproxy-reloads/ for solutions that are even better than this one.
- jsprogrammer 11y agoWhy shouldn't HAProxy have an API for updating its configuration while running?
- vidarh 11y agoIt should, and it allows you to do some changes over a socket, but allowing complete updates with the current haproxy architecture is a lot of work and isn't there yet.