Hi, I'm proxying UDP, with no healthchecks and both client_ and backend_idle_timeouts set to 1m. I have two servers configured statically. I notice if I stop the service on one of them, gobetween notices within a few seconds and starts sending the traffic to the other one. The error logged is
(udp/server/session): Failed to read from backend: read udp 192.168.60.1:35660->192.168.60.12:6121: read: connection refused. This is great, I'm just wondering what specifically the failover behavior is. Since it's UDP there's no session or ACKs, etc to keep track of at the transport layer, and I'm not sure what connection refused means in the context of UDP. Thanks.
which prompts gobetween to close that session and try again on the other server
I'm not seeing actual roundrobin UDP
with a stream- it looks like it picks a backend host and delivers all packets to it
I assume it thinks it is a stream and delivers to the first host it picks.
I'm wondering if I can force it to cycle destinations to do true load balancing rather than trying to some kind of session-based load balancing..
BTW, have checked the wiki at: https://github.com/yyyar/gobetween/wiki
in case you have more queries, it would be good to know about your setup details :)
in case you have more queries, it would be good to know about your setup details :)
Hi. Has anyone had success with AMQP behind gobetween? I've tried using gb directly over three rabbitmq servers in hash1 mode, but I'm getting odd errors in response. I can proxy with amqproxy fine, but that solution doesn't handle multiple backends, so I thought I'd bridge the two to get some sort of HA to my proxy (running 3 nodes within a kubernetes cluster)
Hey guys, I've just downloaded the latest version of gobetween. I then pretty much used the default provided config to test load balancing to a couple websphere app instances I already had running. however, when hitting the load balancer I am actually redirected to the particular WAS instance URL, which is obviously not what I wanted. I couldn't find anything in the documentation in relation to this, is that expected behaviour?
FYI, I found out what the problem was. Websphere itself was redirecting. This article is what I had to do to fix it https://www-01.ibm.com/support/docview.wss?uid=swg21569667
Hi Guys, has anyone had any luck with the exec health check in windows? I have a really simple scenario which is more failover then load balance, I'm using a batch file and curl to check which site is active but whatever "expected output" I put in the config and the bat file it's always "unexpected" - batch file copies the example in the documentation albeit with my value
@echo off
curl -o nul --silent --head --fail -Is http://%1/
IF /I "%ERRORLEVEL%" NEQ "0" (
echo|set /p Dummy="DOWN"
)ELSE (
echo|set /p Dummy="UP"
)
and the healthcheck is
kind = "exec"
interval = "10s"
timeout = "2s"
exec_command ="C:\myscript\gobetween\healthcheck\exec_healthcheck.bat" # (required) command to execute
exec_expected_positive_output = "UP" # (required) expected output of command in case of success
exec_expected_negative_output = "DOWN" # (required) expected output of command in case of failure
status is always
@echo off
curl -o nul --silent --head --fail -Is http://%1/
IF /I "%ERRORLEVEL%" NEQ "0" (
echo|set /p Dummy="DOWN"
)ELSE (
echo|set /p Dummy="UP"
)
and the healthcheck is
kind = "exec"
interval = "10s"
timeout = "2s"
exec_command ="C:\myscript\gobetween\healthcheck\exec_healthcheck.bat" # (required) command to execute
exec_expected_positive_output = "UP" # (required) expected output of command in case of success
exec_expected_negative_output = "DOWN" # (required) expected output of command in case of failure
status is always
so it's working and it's seeing it but I can't get the output to match, tried quoting, unquoting etc
why not jist echo "UP" or echo "DOWN" ?
i use "0" and "1", my batch file calls a powershell script which then does a "Write-Host '0' -NoNewLine" or "Write-Host '1' -NoNewLine" depending on the server health
Wait, we got it
your script outputs "UP" with "
including quotation symbols
@JustinButterwo8_twitter
so "UP" does not match with UP :-)
So somehow remove quotes around UP and DOWN from healthcheck script output and it should fix the issue :-)
and echo DOWN
Hi,
I'm a newbie. This's first time I use gobetween. I want to config sample Load balancing for two webservice, which request come from domain: mydomain.com to IP of server installed gobetween. But I can't see any documentation for it. I'm setting on my windows server 2012. I really need your help, are there anyone who can help me?
Thanks so much!
I'm a newbie. This's first time I use gobetween. I want to config sample Load balancing for two webservice, which request come from domain: mydomain.com to IP of server installed gobetween. But I can't see any documentation for it. I'm setting on my windows server 2012. I really need your help, are there anyone who can help me?
Thanks so much!
This is my config.
[servers.gklb]
bind = "localhost:3000"
protocol = "tcp"
balance = "weight"
max_connections = 10000
client_idle_timeout = "10m"
backend_idle_timeout = "10m"
backend_connection_timeout = "2s"
[servers.gklb.discovery]
kind = "static"
static_list = [
"IP server 1:80 weight=1",
"IP server 2:80 weight=1",
]
[servers.gklb.healthcheck]
fails = 1
passes = 1
interval = "2s"
timeout="1s"
kind = "ping"
ping_timeout_duration = "500ms"
And I want [bind = "localhost:3000"] is my domain which is used by end user on your browser.
I'm familiar with Haproxy and looking to replace most of the load balancing for tcp and udp ports to a Graylog cluster.