Service advertising behavior differs on Windows vs. Linux

[webcoyote]

The Windows and Linux versions of consul behave differently: the Windows version is >not< advertising a service in catalog/services though it is listed in agent/services, whereas it >is< advertising in both on Linux. This issue is 100% reproducible on my computers, not random behavior.

The key difference in the output is:

Linux: ~$ time curl localhost:8500/v1/catalog/services
{"consul":[],"test":["blah"]}

Windows: ~$ time curl localhost:8500/v1/catalog/services
{"consul":null}

Note also the long time it takes for Windows to respond to queries compared to Linux.

In command shell 1, run consul

mkdir -p ~/consul/test/consul.d
cd ~/consul/test
echo '{"service":{"name": "test","tags": ["blah"],"port": 9001,"check":{"name": "Test status","ttl": "15s"}}}' > ./consul.d/test.json
rm ./data -rf ; GOMAXPROCS=2 consul agent -bootstrap -server -data-dir=./data -bind=127.0.0.1 -config-dir=./consul.d -log-level=trace -node="$OSTYPE"

In command shell 2, run tests

time curl localhost:8500/v1/agent/services
time curl localhost:8500/v1/agent/checks
time curl localhost:8500/v1/catalog/services

Linux test output

~$ time curl localhost:8500/v1/agent/services
{"test":{"ID":"test","Service":"test","Tags":["blah"],"Port":9001}}

real  0m0.005s
user  0m0.000s
sys 0m0.000s
pat@vm88-UL:~$ time curl localhost:8500/v1/agent/checks
{"service:test":{"Node":"linux-gnu","CheckID":"service:test","Name":"Service 'test' check","Status":"unknown","Notes":"","Output":"","ServiceID":"test","ServiceName":"test"}}

real  0m0.004s
user  0m0.000s
sys 0m0.000s
pat@vm88-UL:~$ time curl localhost:8500/v1/catalog/services
{"consul":[],"test":["blah"]}

real  0m0.007s
user  0m0.000s
sys 0m0.004s
~$ 

Windows test output

~$ time curl localhost:8500/v1/agent/services
{"test":{"ID":"test","Service":"test","Tags":["blah"],"Port":9001}}

real    0m2.039s
user    0m0.015s
sys     0m0.000s
~$ time curl localhost:8500/v1/agent/checks
{"service:test":{"Node":"msys","CheckID":"service:test","Name":"Service 'test' check","Status":"unknown","Notes":"Initializing","ServiceID":"test","ServiceName":"test"}}

real    0m2.038s
user    0m0.016s
sys     0m0.000s
~$ time curl localhost:8500/v1/catalog/services
{"consul":null}

real    0m1.028s
user    0m0.015s
sys     0m0.000s
~$

Linux consul output

~/consul/test$ rm ./data -rf ; GOMAXPROCS=2 consul agent -bootstrap -server -data-dir=./data -bind=127.0.0.1 -config-dir=./consul.d -log-level=trace -node="$OSTYPE"
==> WARNING: Bootstrap mode enabled! Do not enable unless necessary
==> Starting Consul agent...
==> Starting Consul agent RPC...
==> Consul agent running!
       Node name: 'linux-gnu'
      Datacenter: 'dc1'
          Server: true (bootstrap: true)
     Client Addr: 127.0.0.1 (HTTP: 8500, DNS: 8600, RPC: 8400)
    Cluster Addr: 127.0.0.1 (LAN: 8301, WAN: 8302)

==> Log data will now stream in as it occurs:

    2014/05/28 10:24:36 [INFO] raft: Node at 127.0.0.1:8300 [Follower] entering Follower state
    2014/05/28 10:24:36 [INFO] serf: EventMemberJoin: linux-gnu 127.0.0.1
    2014/05/28 10:24:36 [INFO] serf: EventMemberJoin: linux-gnu.dc1 127.0.0.1
    2014/05/28 10:24:36 [INFO] consul: adding server for datacenter: dc1, addr: 127.0.0.1:8300
    2014/05/28 10:24:36 [DEBUG] dns: request for {_test.consul. 255 1} (95.541us)
    2014/05/28 10:24:36 [DEBUG] dns: request for {_test.consul. 255 1} (144.149us)
    2014/05/28 10:24:36 [ERR] agent: failed to sync remote state: rpc error: No cluster leader
    2014/05/28 10:24:37 [WARN] raft: Heartbeat timeout reached, starting election
    2014/05/28 10:24:37 [INFO] raft: Node at 127.0.0.1:8300 [Candidate] entering Candidate state
    2014/05/28 10:24:37 [DEBUG] raft: Votes needed: 1
    2014/05/28 10:24:37 [DEBUG] raft: Vote granted. Tally: 1
    2014/05/28 10:24:37 [INFO] raft: Election won. Tally: 1
    2014/05/28 10:24:37 [INFO] raft: Node at 127.0.0.1:8300 [Leader] entering Leader state
    2014/05/28 10:24:37 [INFO] consul: cluster leadership acquired
    2014/05/28 10:24:37 [INFO] consul: New leader elected: linux-gnu
    2014/05/28 10:24:37 [INFO] consul: member 'linux-gnu' joined, marking health alive
    2014/05/28 10:24:39 [INFO] agent: Synced service 'test'
    2014/05/28 10:24:39 [INFO] agent: Synced check 'service:test'
    2014/05/28 10:24:45 [DEBUG] http: Request /v1/agent/services (94.144us)
    2014/05/28 10:24:45 [DEBUG] http: Request /v1/agent/checks (39.39us)
    2014/05/28 10:24:46 [DEBUG] http: Request /v1/catalog/services (1.257678ms)
    2014/05/28 10:24:51 [WARN] Check 'service:test' missed TTL, is now critical
    2014/05/28 10:24:51 [DEBUG] agent: Service 'test' in sync
    2014/05/28 10:24:51 [INFO] agent: Synced check 'service:test'

Windows consul output

~/consul/test$ rm ./data -rf ; GOMAXPROCS=2 consul agent -bootstrap -server -data-dir=./data -bind=127.0.0.1 -config-dir=./consul.d -log-level=trace -node="$OSTYPE"
==> WARNING: Bootstrap mode enabled! Do not enable unless necessary
==> WARNING: Windows is not recommended as a Consul server. Do not use in production.
==> Starting Consul agent...
==> Starting Consul agent RPC...
==> Consul agent running!
       Node name: 'msys'
      Datacenter: 'dc1'
          Server: true (bootstrap: true)
     Client Addr: 127.0.0.1 (HTTP: 8500, DNS: 8600, RPC: 8400)
    Cluster Addr: 127.0.0.1 (LAN: 8301, WAN: 8302)

==> Log data will now stream in as it occurs:

    2014/05/28 10:22:43 [INFO] raft: Node at 127.0.0.1:8300 [Follower] entering Follower state
    2014/05/28 10:22:43 [INFO] serf: EventMemberJoin: msys 127.0.0.1
    2014/05/28 10:22:43 [INFO] serf: EventMemberJoin: msys 127.0.0.1
    2014/05/28 10:22:43 [INFO] consul: adding server for datacenter: dc1, addr: 127.0.0.1:8300
    2014/05/28 10:22:43 [DEBUG] dns: request for {_test.consul. 255 1} (0)
    2014/05/28 10:22:43 [DEBUG] dns: request for {_test.consul. 255 1} (0)
    2014/05/28 10:22:43 [ERR] agent: failed to sync remote state: rpc error: No cluster leader
    2014/05/28 10:22:43 [ERR] agent: failed to sync remote state: rpc error: No cluster leader
    2014/05/28 10:22:44 [WARN] raft: Heartbeat timeout reached, starting election
    2014/05/28 10:22:44 [INFO] raft: Node at 127.0.0.1:8300 [Candidate] entering Candidate state
    2014/05/28 10:22:44 [DEBUG] raft: Votes needed: 1
    2014/05/28 10:22:44 [DEBUG] raft: Vote granted. Tally: 1
    2014/05/28 10:22:44 [INFO] raft: Election won. Tally: 1
    2014/05/28 10:22:44 [INFO] raft: Node at 127.0.0.1:8300 [Leader] entering Leader state
    2014/05/28 10:22:44 [INFO] consul: cluster leadership acquired
    2014/05/28 10:22:44 [INFO] consul: New leader elected: msys
    2014/05/28 10:22:44 [INFO] consul: member 'msys' joined, marking health alive
    2014/05/28 10:22:51 [DEBUG] http: Request /v1/agent/services (0)
    2014/05/28 10:22:53 [DEBUG] http: Request /v1/agent/checks (0)
    2014/05/28 10:22:54 [DEBUG] http: Request /v1/catalog/services (500.1us)
    2014/05/28 10:22:58 [WARN] Check 'service:test' missed TTL, is now critical

[webcoyote]

The problem may be related to this:

Linux

~$ curl -X PUT http://localhost:8500/v1/agent/service/register -d '{"name": "test2","port": 9002,"check"
:{"name": "Test2 status","ttl": "15s"}}'
==> no output

Windows

~$ curl -X PUT http://localhost:8500/v1/agent/service/register -d '{"name": "test2","port": 9002,"check"
:{"name": "Test2 status","ttl": "15s"}}'
==> Request decode failed: json: cannot unmarshal string into Go value of type time.Duration

[armon]

Are you running the latest version for both? The decoding error was fixed in 0.2 I believe.

[webcoyote]

I thought I was running latest but it turns out that the Windows binary built for v0.2.1 contains the wrong version:
https://dl.bintray.com/mitchellh/consul/0.2.1_windows_386.zip <<< actually v0.1.0
Edit: It turns out that 0.2.0_windows_386.zip is also actually version v0.1.0

[armon]

Oh yikes! I will try to get a new binary uploaded today. I'll bug @mitchellh since he has a windows machine.

[webcoyote]

I compiled consul from source and that version functions properly on Windows.

[armon]

@webcoyote The download for 0.2.1 should be fixed now! Thanks for reporting!