[Fix] Service tags not added to health checks

[yfouquet]

Since commit 9685bdc, service tags are added to the health checks.
Otherwise, when adding a service, tags are not added to its check.

In updateSyncState, we compare the checks of the local agent with the checks of the catalog.
It appears that the service tags are different (missing in one case), and so the check is synchronized.
That increase the ModifyIndex periodically when nothing changes.

Fixed it by adding serviceTags to the check.

Note that the issue appeared in version 0.8.2.
Looks related to issue #3259.

[yfouquet]

Hi @slackpad @magiconair,
I really think that it would fix some issues for blocking query users.
Could you please have a look at it?

[yfouquet]

A simple way to see the issue is to compare checks seen by the agent and health endpoints.
In ../health/service/<my_service>, the ServiceTags of some checks are not empty.
In .../agent/checks, the ServiceTags are empty for the same checkId

[yfouquet]

Hi @slackpad,
I have adjusted the documentation regarding ServiceTags in health checks.
Please tell me what can be done to help you handle this PR.

[yfouquet]

Hi,
We have deployed consul agents with this patch in our datacenters.
The result is that long-polling requests can now last 10 minutes instead of a few seconds (or milliseconds for some services).
The following graph shows the long-polling request duration and count on a service with 500 instances inside the local datacenter.

[slackpad]

Thank you so much for finding this and submitting a PR!

[magiconair]

oops...