BUG/MEDIUM: checks: Use the mux protocol specified on the server line
First, when a server health check is initialized, it inherits the mux protocol
from the server if it is not already specified. Because there is no option to
specify the mux protocol for the checks, it is always inherited from the server
for now.
Then, if the connect rule is configured to use the server options, the mux
protocol of the check is used, if defined. Of course, if a mux protocol is
already defined for the connect rule, it is used in priority. But for now, it is
not possible.
Thus, if a server is configured to use, for instance, the h2 protocol, it is
possible to do the same for the health-checks.
No backport needed.
diff --git a/src/checks.c b/src/checks.c
index 7dff7e9..0bd84c0 100644
--- a/src/checks.c
+++ b/src/checks.c
@@ -1743,12 +1743,13 @@
* check (email alerts) or if there is a mux proto specified or if there
* is no alpn.
*/
- if (!s || connect->mux_proto || check->mux_proto || (!connect->alpn && !check->alpn_str)) {
+ if (!s || ((connect->options & TCPCHK_OPT_DEFAULT_CONNECT) && check->mux_proto) ||
+ connect->mux_proto || (!connect->alpn && !check->alpn_str)) {
const struct mux_ops *mux_ops;
if (connect->mux_proto)
mux_ops = connect->mux_proto->mux;
- else if (check->mux_proto)
+ else if ((connect->options & TCPCHK_OPT_DEFAULT_CONNECT) && check->mux_proto)
mux_ops = check->mux_proto->mux;
else {
int mode = ((check->tcpcheck_rules->flags & TCPCHK_RULES_PROTO_CHK) == TCPCHK_RULES_HTTP_CHK
@@ -4724,6 +4725,12 @@
srv->check.send_proxy |= (srv->pp_opts);
}
+ /* Inherit the mux protocol from the server if not already defined for
+ * the check
+ */
+ if (srv->mux_proto && !srv->check.mux_proto)
+ srv->check.mux_proto = srv->mux_proto;
+
/* validate <srv> server health-check settings */
/* We need at least a service port, a check port or the first tcp-check