Better healthcheck check about tapalcatl

[rmarianski]

I like health checks to be very simple, and would prefer the first one. And within that, would prefer to request an object instead of listing the bucket, because that's what the service does. We could just make a file at the root of the bucket like you mentioned, eg /_healthcheck and have all services that depend on s3 check that.

[nvkelso]

I also prefer the single file. I'm fine with new health check file, or we could check that the tilejson exists (which has gone missing in the past).

…

On Mar 1, 2017, at 07:33, Robert Marianski ***@***.***> wrote: I like health checks to be very simple, and would prefer the first one. And within that, would prefer to request an object instead of listing the bucket, because that's what the service does. We could just make a file at the root of the bucket like you mentioned, eg /_healthcheck and have all services that depend on s3 check that. ― You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[iandees]

Yep, I agree that we should check for a tile. I'd say the 0/0/0 tile rather than a file we have to remember to add.

Are there other services that tapalcatl uses that we should also check in tapalcatl's healthcheck?

[rmarianski]

I'd say the 0/0/0 tile rather than a file we have to remember to add.

I still prefer the synthetic file. We'd only have to add the synthetic file once, and never again for subsequent deployments. Checking 0/0/0 would mean that we would be unable to go live with a new date prefix unless we were sure that 0/0/0 was generated first.

Are there other services that tapalcatl uses that we should also check in tapalcatl's healthcheck?

Don't think so. We send metrics over udp, but that shouldn't be involved with the health check.

[nvkelso]

Would the synthetic health check file need to be added per deploy into each new date prefixed bucket?

[rmarianski]

Would the synthetic health check file need to be added per deploy into each new date prefixed bucket?

No. See #23 (comment)

[nvkelso]

Because the health check would talk directly to S3 and not to the tile.mapzen.com endpoint?

…

On Wed, Mar 1, 2017 at 8:48 AM, Robert Marianski ***@***.***> wrote: Would the synthetic health check file need to be added per deploy into each new date prefixed bucket? No. See #23 (comment) <#23 (comment)> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#23 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA0EO-3Ku94vikSVwOOHC_e-U1FXyU1mks5rhaFHgaJpZM4MPfYl> .

[rmarianski]

Because the health check would talk directly to S3

Yes, exactly.