Release 1.0.0

[2opremio]

@davkal @foot @ekimekim @bowenli @jpellizzari @tomwilkie @jml @abuehrle @pidster @errordeveloper This is a major release, please chip in by testing this release branch in the scenarios below and tick off what you tested afterwards (including your name at the end).

EDIT: To make things easier to test I have pushed the image weaveworks/scope:release-1.0-cd1e06b

Thanks!

• Start a 2 node cluster and leave a UI connected, watch for performance regression (Done by @2opremio)
• Run a Scope cluster on ECS (Done by @2opremio)
• Run Scope on https://github.com/microservices-demo/microservices-demo/
• Run Scope on Docker For Mac
• Start and stop Scope in standalone mode
  • Linux
  • docker-machine
• Connect probes from release candidate to Weave Cloud
  • Local scope in probe-only mode connects to Weave Cloud and renders correctly
    • dev
    • prod
  • Controls appear for containers and function correctly (Done by @2opremio)
    • dev
    • prod
  • No new errors in logging or telemetry
    • dev
    • prod

Things to think about

• Kubernetes Integration
• CPU and memory usage

[jml]

Won't have time before the release, sorry.

[pidster]

I've tested on

• SockShop locally (docker compose) + local Scope
• SockShop locally (docker compose) + Weave Cloud (dev)
• SockShop locally (docker compose) + Weave Cloud (prod)
• D4M

[2opremio]

Thanks @pidster. Can you tick it off from the check list?

[pidster]

I've ticked off the things I've tested

[abuehrle]

@fons I tested the latest on my ubuntu host and probes are showing up in Cloud.

and it does start and stop.

It allowed me to include a command twice, i.e.: scope launch scope launch token=<xxxx>

(this is why the probes didn't show up at first, but probably shouldn't have let me include the command in here twice.)

[rade]

It allowed me to include a command twice, i.e.: weave launch weave launch token=<xxxx>

@abuehrle you mean scope launch, right?

[abuehrle]

Yes, I meant scope launch. I will log this issue for you. Probably a whole lot more tests we could do there as well.

[pidster]

I've tested docker-machine on MacOS.

[2opremio]

Start a 2 node cluster and leave a UI connected, watch for performance regression

I actually updated the probes in the dev cluster https://github.com/weaveworks/service-conf/pull/324 and I observed they are consuming ~70% CPU which is pretty horrible but I don't think it's a regression. I am getting a CPU profile just in case.

[2opremio]

Here it goes, nothing new is spiking up. However, conntrack's XML parser seems to be consuming too much CPU. I had noticed how the parser was less than optimal, but 42% CPU is big enough to consider switching to another parser.

[pidster]

@2opremio is it worth replacing the conntrack parser if there's an alternative to conntrack in the pipe?

[2opremio]

@pidster Please comment in the relevant issue #1991

This issue is for the release (and we are not going to replace the parser before then)