These are chat archives for ethersphere/orange-lounge

19th
Apr 2018
Anton Evangelatov
@nonsense
Apr 19 07:27
@holisticode i am around today
lash
@nolash
Apr 19 16:18
Should this file be removed? It doesn't seem to be in use. swarm/network/simulations/overlay.go
Viktor Trón
@zelig
Apr 19 16:53
nooo
it is in use for visualisation :)
backend to visualiser is go run ...overlay.go
lash
@nolash
Apr 19 16:54
Ok, file is broken, though.
lash
@nolash
Apr 19 17:12
this is great for now for this PR abut i want to replace this test with one using janos's dynamic framework with a global health check
@zelig @janos what is this framework?
holisticode
@holisticode
Apr 19 17:46
Should this file be removed? It doesn't seem to be in use. swarm/network/simulations/overlay.go
As @zelig said, this is the actual backend for the visualisation
the problem with it is that it doesn't have any test file for that visualisation itself, it's only a main so it easily gets missed when people change code and it breaks
I thought there was a ticket to add that so at least it breaks when tests are run
lash
@nolash
Apr 19 17:48
yeah, there probably should be.
holisticode
@holisticode
Apr 19 17:53
@nonsense looking at metrics it actually doesn't look too bad with CPU load and memory usage?
Or am I misinterpreting things - I was expecting a bit more "sweating"
Viktor Trón
@zelig
Apr 19 18:28
that is good news @holisticode but lets wait for people bombarding the gateways
holisticode
@holisticode
Apr 19 18:28
sure
Anton Evangelatov
@nonsense
Apr 19 18:33
well the new release is certainly better than the previously deployed version
well done @janos and everyone else involved :)
especially the CPU profile looks much better
holisticode
@holisticode
Apr 19 18:34
much better
even when we were uploading 50-80MB files, no significant spikes
Anton Evangelatov
@nonsense
Apr 19 18:35
i will probably trigger a continuous smoke test run at some point
holisticode
@holisticode
Apr 19 18:35
well more load for sure, but nothing to worry about
yes!
Anton Evangelatov
@nonsense
Apr 19 18:35
@holisticode don’t forget that those machines are quite beafy
holisticode
@holisticode
Apr 19 18:35
that is what we should use the test cluster for
Anton Evangelatov
@nonsense
Apr 19 18:35
even though we run 4 containers per machine
holisticode
@holisticode
Apr 19 18:35
beafy how?
Anton Evangelatov
@nonsense
Apr 19 18:36
they are 28GB RAM and 4vCPU
ok, not so beafy, but still 4 CPUs
holisticode
@holisticode
Apr 19 18:37
and every docker image has how much memory?
Anton Evangelatov
@nonsense
Apr 19 18:37
CONTAINER           CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
048c40aec166        18.38%              1.016GiB / 5GiB       20.33%              3.54GB / 2.55GB     0B / 17.6GB         22
803bd754aa3f        18.22%              1.026GiB / 5GiB       20.52%              3.43GB / 2.81GB     8.19kB / 17.5GB     23
33911a21ddf7        13.39%              1.061GiB / 5GiB       21.21%              1.57GB / 14.6GB     4.1kB / 13.4GB      24
d79a0374bae7
run docker stats on one of them
they are limited to 5GB for RAM
holisticode
@holisticode
Apr 19 18:38
5GB
Anton Evangelatov
@nonsense
Apr 19 18:38
i don’t think i have added a limit on CPU
holisticode
@holisticode
Apr 19 18:38
well that's not mindblowingly beafy either, so I think I join to the applause for everyone contributing to this improvements!
will we finally have an open test cluster as well?
Brian Soule
@briansoule
Apr 19 18:39
👏
Anton Evangelatov
@nonsense
Apr 19 18:40
yes, true ;) just wanted to make it clear we are not running multiple swarm containers on a small instance or smth
holisticode
@holisticode
Apr 19 18:43
Would be nice to interact with an open test cluster, upload / retrieve, and observe behavior over time....
Anton Evangelatov
@nonsense
Apr 19 18:48
yes, something i will be improving over the coming days
holisticode
@holisticode
Apr 19 18:49
excellent @nonsense
Anton Evangelatov
@nonsense
Apr 19 18:49
i am deploying the same version to it now
however it currently has no bootnode, so i don’t think it will be easy for any of the containers to discover each other
holisticode
@holisticode
Apr 19 18:50
swarmtest1 has a boot node though
Anton Evangelatov
@nonsense
Apr 19 18:50
this is the closed test cluster
i setup an open one (that is no behind firewall)
holisticode
@holisticode
Apr 19 18:51
aah gotcha
Anton Evangelatov
@nonsense
Apr 19 18:52
048c40aec166        swarm               "/bin/smell-baron ..."   5 hours ago         Up 5 hours          6060/tcp, 10.0.1.5:30401->30401/tcp, 10.0.1.5:30401->30401/udp, 0.0.0.0:8503->8500/tcp   swarm_8503
notice that the ports are listening on the private ip
holisticode
@holisticode
Apr 19 18:52
yep
Anton Evangelatov
@nonsense
Apr 19 18:52
on the open test cluster and prod:
swarmopen@swarmopen-1:~$ docker ps
CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS                                                                        NAMES
8dde316156a5        swarm               "/bin/smell-baron ..."   30 seconds ago      Up 29 seconds       6060/tcp, 8500/tcp, 0.0.0.0:30401->30401/tcp, 0.0.0.0:30401->30401/udp       swarm_30401
holisticode
@holisticode
Apr 19 18:52
:+1: