745299 - all inside-the-build requests for HG should use hg-internal.dmz.scl3.m.c instead of hg.m.o

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P3)

Description

[Mike Taylor [:bear]]

currently we are using only hg.m.o and that means we share the same paths and constraints as public facing HG access

Comment 1

[Chris AtLee [:catlee]]

bkero: should we also be using 'hg clone --uncompressed' as much as possible with this move?

Comment 2

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

from irc:

1) this seems to be related to yesterday's hg migration, but unclear.

2) is "use hg-internal.m.o instead of hg.m.o" to be done as well as (or instead of?) using the two hg mirrors?

Comment 3

[Ben Kero [:bkero]]

catlee: hg clone --uncompressed should be used (using the 'hg checkout --uncompressed http:// interface) only when using the hg-internal.dmz.scl3.mozilla.com address.

joduinn: it was originally going to be related to the hg migration, although I did not get clearance from opsec in time for these both to happen concurrently.

The hg mirrors can be used in case of network connectivity problems between scl1 and scl3.  We should re-evaluate if the hg mirrors in scl1 are necessary.

Comment 4

[Chris AtLee [:catlee]]

Attachment: add hg-internal as the first mirror

mirrors are tried in order, so in theory this will try the new hg-internal mirror first before falling back to the old mirror.

Comment 5

[Chris AtLee [:catlee]]

deployed in today's reconfig