On 11/6/24 10:39 AM, Nick Couchman wrote:
On Wed, Nov 6, 2024 at 1:30 PM Steven Barnhart <[email protected]>
wrote:
This has been our experience at least lately as well. The connection groups
aren’t balancing.
To say "the connection groups aren't balancing" is not accurate - they do
balance, but there is currently no mechanism for checking the state of a
host to see if it is up before assigning that host to a user and attempting
to connect.
Steve
On Wed, Nov 6, 2024 at 1:28 PM <[email protected]> wrote:
Hello everyone.
My scenario.
I have a Connection group, set up as load balance, with 10 machines. Due
to
pay-per-use issues, only 5 are turned on at any given time.
When a user connects, he/she receives a machine that is turned off.
At this point, he/she enters an error loop.
My question: is there any configuration that can set that when connecting
to a machine with this error loop, it is redirected to the next machine
in
the connection group, until it finds a machine that is available.
At the moment, no. But there should be :-).
There is failover functionality built into the connection groups that
will automatically switch to a different connection in the group if the
connection attempt fails, but that functionality depends on getting an
explicit failure that indicates a different connection should be tried.
In the case that several servers are entirely offline, then there likely
is no explicit failure except a TCP connection timeout. By the time that
a timeout occurs, the overall connection process may have already taken
so long that it gets abandoned before another connection is tried.
If this is what's happening, then a workaround would be to try reducing
the connection timeout via the timeout connection parameter introduced
on the "staging/1.6.0" branch. Ideally, the network infrastructure
between the Guacamole server and the destination server should respond
directly with "host unreachable" when a TCP connection attempt is made
to an offline host, avoiding any delay.
- Mike
