1202513 - Route traffic from AWS slaves to new upload hosts via IGW

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Nick Thomas [:nthomas] (UTC+12)]

There are 6 new hosts to replace the upload role of stage.m.o:

upload.tbirdbld.productdelivery.stage.mozaws.net		52.3.217.123
upload.ffxbld.productdelivery.stage.mozaws.net		52.1.123.79
upload.trybld.productdelivery.stage.mozaws.net		52.0.28.69

upload.tbirdbld.productdelivery.prod.mozaws.net		52.88.134.149
upload.ffxbld.productdelivery.prod.mozaws.net		52.11.216.122
upload.trybld.productdelivery.prod.mozaws.net		52.27.28.242

(and maybe b2gbld as well, but that's unresolved right now). I would like to route traffic from AWS instances directly to those AWS hosts via the IGW, rather than down the tunnel to SCL3 and then back to AWS.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

I see we don't currently route upload traffic to stage.m.o via IGW, could you remind me why that is ? Perhaps this is WONTFIX.

Comment 2

[Chris AtLee [:catlee]]

I don't recall, sorry.

Comment 3

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: Pull request

Comment 4

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: Log for routing table update

This adds the new hosts, and also
* ftp-ssl.m.c becomes 54.192.117.66 since oremj moved that to cloudfront as part of the s3 migration; luckily we get the same endpoint running the script in scl3 as the actual instances do in use1/usw2. We remove 63.245.215.126, 63.245.215.127, and 63.245.215.
* 192.81.221.125/32, 78.46.92.172/32, and 88.198.22.7/32 are removed. I suspect are hostedgraphite.com hosts, which have a lot of IPs in round robins. In retrospect I should have left them alone
* I didn't delete any tables not in the config (proxxy etc)

Comment 5

[Nick Thomas [:nthomas] (UTC+12)]

Spot checks of instances and masters in use1+usw2 look good, they can reach upload.tbirdbld.productdelivery.prod.mozaws.net.