[Varnish] #373: Varnish fails to reload backend/directory
Varnish
varnish-bugs at projects.linpro.no
Tue Nov 11 10:00:20 CET 2008
#373: Varnish fails to reload backend/directory
----------------------+-----------------------------------------------------
Reporter: anders | Owner: phk
Type: defect | Status: new
Priority: normal | Milestone:
Component: varnishd | Version: trunk
Severity: normal | Keywords:
----------------------+-----------------------------------------------------
When reloading VCL by using vcl.load and vcl.use, Varnish fails to update
which hosts to probe.
Given this VCL:
{{{
backend kart3d1 {
.host = "192.168.41.134";
.port = "80";
.probe = {
.request =
"GET /varnish-lb-check.html HTTP/1.1"
"Host: foo.bars.no"
"Connection: close";
.timeout = 5s;
.window = 8;
.threshold = 3;
.interval = 2s;
}
}
backend kart3d2 {
.host = "192.168.41.135";
.port = "80";
.probe = {
.request =
"GET /varnish-lb-check.html HTTP/1.1"
"Host: foo.bars.no"
"Connection: close";
.timeout = 5s;
.window = 8;
.threshold = 3;
.interval = 2s;
}
}
director kart3dlb random {
{
.backend = kart3d1;
.weight = 100;
}
{
.backend = kart3d2;
.weight = 100;
}
}
}}}
If one backend is commented out (as backend, and in the directory), and
VCL is reloaded, Varnish still polls it. If the host header is updated, a
full restart of Vartnish is also necessary.
Conclusion: it seems Varnish is unable to update backend/director runtime,
without a restart. It would be nice to have this fixed, so that cache
servers with a large dataset do not lose their data because of load
balancing changes.
--
Ticket URL: <http://varnish.projects.linpro.no/ticket/373>
Varnish <http://varnish.projects.linpro.no/>
The Varnish HTTP Accelerator
More information about the varnish-bugs
mailing list