Perf server initial documentation; mostly placeholder now.

perf/server/README.md

+SkiaPerf Server
+===============
+
+Reads Skia performance data from databases and serves interactive dashboards for easy exploration and annotations.
+
+Server Setup
+============
+

[rmistry, 2014/06/04 13:59:18]

We can instead point from here to a README file in compute_engine_scripts that
will contain information on how to create, setup and delete this server
referring to the corresponding scripts.

[benchen, 2014/06/04 15:23:32]

Done.

+Create a GCE instance:
+
+    gcutil --project=google.com:skia-buildbots addinstance skia-perf-b \
+      --zone=us-central2-b --external_ip_address=108.170.220.208 \
+      --service_account=default \
+      --service_account_scopes="https://www.googleapis.com/auth/devstorage.full_control" \
+      --network=default --machine_type=n1-standard-1 --image=backports-debian-7-wheezy-v20140331 \
+      --persistent_boot_disk

[rmistry, 2014/06/04 13:59:18]

This command is going to be the one that will be in vm_create_server.sh but the
constants will be used from
https://skia.googlesource.com/buildbot/+/master/compute_engine_scripts/buildb...

[benchen, 2014/06/04 15:23:32]

Got it. Is it fine to share and put new constants under buildbots, or we should
have a new vm_config?
On 2014/06/04 13:59:18, rmistry wrote:

[rmistry, 2014/06/04 15:26:46]

Should be fine to share vm_config and put new constants there.

+
+Make sure port 80 is accessible externally for the above instance.
+
+SSH into the instance:
+
+    gcutil --project=google.com:skia-buildbots ssh --ssh_user=default skia-perf-b

[rmistry, 2014/06/04 13:59:18]

We should try to avoid hardcoding these values because if we migrate due to a
PCR then this may become skia-perf-a.

[benchen, 2014/06/04 15:23:32]

I see, will figure out how it is done currently.
On 2014/06/04 13:59:18, rmistry wrote:

+
+Setup scripts can be found in the buildbot repo under compute_engine_scripts/perfserver.