You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now the README gives the impression that in order to achieve seamless reloads you need to enable the stats socket and include expose-fd listeners:
To achieve seamless reloads it is required to use expose-fd listeners and socket transfers which are not enabled by default.
In master-worker mode, this is not required anymore, the listeners will be passed using the internal socketpairs between the master and the workers.
Since all versions of the entrypoint.sh script automatically enables master-worker mode via the -W flag it seems to me that for at least 2.6 and up you actually don't need to have the stats socket enabled at all for seamless reloads to work?
All your points are valid and I can confirm fixing README is in the queue to be done really soon. My apologies for being late on this (and being late with the response), but we are working on a few key projects at the moment..
Hello,
Right now the README gives the impression that in order to achieve seamless reloads you need to enable the stats socket and include
expose-fd listeners
:However, looking at e.g. https://docs.haproxy.org/2.6/configuration.html#5.1-expose-fd%20listeners it states:
Since all versions of the
entrypoint.sh
script automatically enables master-worker mode via the-W
flag it seems to me that for at least 2.6 and up you actually don't need to have the stats socket enabled at all for seamless reloads to work?On a similar note the README also stats that you send
SIGHUP
to reload the configuration, but reading even the older 2.2 version of the docs (https://docs.haproxy.org/2.2/configuration.html#3.1-master-worker) it states:So maybe the selection of signal should be updated as well?
The text was updated successfully, but these errors were encountered: