7bc9ca6c8aecc3347402233a20fec8ce1580b2a2: misc: mass replace readthedocs.org with readthedocs.io
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 16:08:49 -0700 - rev 362605
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
misc: mass replace readthedocs.org with readthedocs.io I just got an email telling me RTD is now serving from readthedocs.io instead of readthedocs.org. They are redirecting the latter to the former. So we might as well update the URLs. While I was here, I snuck in some changes to convert http:// to https://.
d09a58f2334e16ed6be5d1130cf75f82040ab169: ansible/hg-web: unset push_ssl to get rid of "ssl required" message during HTTP push (bug 949153)
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 15:51:38 -0700 - rev 362604
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-web: unset push_ssl to get rid of "ssl required" message during HTTP push (bug 949153) Mercurial only allows pushing via https:// by default because security. When you push to a http:// server, it prints a "ssl required" error message. The problem with our load balancer setup is the load balancer terminates TLS and passes plain text to the origin servers. So Mercurial thinks it is plain text and prints the "ssl required" message even though the client is using https://. We unset push_ssl in the config. The side effect is Mercurial now prints an "authorization required" message because by default Mercurial disallows pushing. This error message is still kind of crappy. But it is less misleading than "ssl required."
4f0dd294022d01942fb597dd8b3453aef5600b1e: ansible/hg-ssh: ensure nslcd.service is enabled and running
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 12:04:51 -0700 - rev 362603
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: ensure nslcd.service is enabled and running
fc6e2aa2c963e7fc8d60dea9627c32867423d5d6: ansible/hg-ssh: always run authconfig
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 12:02:54 -0700 - rev 362602
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: always run authconfig When authconfig is run as a notification, if an ansible deploy updates nslcd.conf and later fails, the notification may never fire and authconfig may never run. Always running authconfig fixes this.
05bd451e47ca44f8d7d4a539c812160977215ea3: ansible/hg-ssh: run pushdataaggregator.service as hg-notifier user/group (bug 1267520)
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 10:57:05 -0700 - rev 362601
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: run pushdataaggregator.service as hg-notifier user/group (bug 1267520)
325354291efaa6f48341aca8351aa6913f719436: ansible/hg-ssh: run pulsenotifier as hg-notifier user/group (bug 1267520)
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 10:56:31 -0700 - rev 362600
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: run pulsenotifier as hg-notifier user/group (bug 1267520)
9f65bc35a80c5eb4ffe3a46902728d43676008ba: ansible/hg-ssh: create hg-notifier user/group (bug 1267520)
Gregory Szorc <gps@mozilla.com> - Wed, 27 Apr 2016 10:56:01 -0700 - rev 362599
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: create hg-notifier user/group (bug 1267520)
c8a45187043a58d9ae000d0532b0b4138890632b: ansible/hg-ssh: install mail-unit-output override as a non-specific override
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 19:20:37 -0700 - rev 362598
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: install mail-unit-output override as a non-specific override systemd doesn't allow you to have N>1 override files for a generic service, it appears. This brings me great sadness. But it isn't a major blow to our desire to send email from failed services.
02916cc0f95d39806e4c8fea5a09fcb9be18fad0: ansible/hg-ssh: add [Install] to hg-bundle-generate.timer
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 19:16:38 -0700 - rev 362597
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: add [Install] to hg-bundle-generate.timer This is needed to get it registered as a timer.
28731a7708ffffc968979097a0c9a9a896254589: ansible/hg-ssh: create directory for override file
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 18:49:44 -0700 - rev 362596
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: create directory for override file Apparently Ansible isn't smart enough to do this on its own.
d5e7089ca219ee94d392f2bac8037414c48128b6: ansible/hg-ssh: fix typo in filename referencing unit override path
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 18:43:40 -0700 - rev 362595
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: fix typo in filename referencing unit override path
7111b9d19232c04bc36f2c0cf81f6949134e2501: ansible/hg-ssh: replace bundle generation CRON with a systemd timer unit
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 18:35:51 -0700 - rev 362594
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: replace bundle generation CRON with a systemd timer unit Systemd can do periodic tasks. It works better than CRON because it will prevent services from running multiple times in parallel and it integrates with the journal, so we will always have timestamped output of CRONs. The service uses our mail-unit-output generic service to send emails on failure. We install a drop-in file to specify where output should go. MozReview-Commit-ID: 7VFq7fZ8Zta
2699a5aeb550c655d6f93054f8f4d3cf133185f0: ansible: install systemd-mail-unit-output files
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 18:30:24 -0700 - rev 362593
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible: install systemd-mail-unit-output files Only on hg-ssh for now. MozReview-Commit-ID: GshOuFZR9L1
067e67ec6c6e4588816b18be2fd058518c8ca14a: ansible: add a generic mechanism to email systemd unit output
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 18:18:57 -0700 - rev 362592
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible: add a generic mechanism to email systemd unit output As we move CRONs to systemd, we need a mechanism to alert people when a CRON has failed. There isn't a built-in mechanism in systemd to do this. What the Internet suggests is you create a generic service unit which executes something that sends a mail. You then use OnFailure in your original unit file to invoke the generic failure service. This commit implements such a system. Our generic service invokes a Python script which invokes journalctl to read the last few minutes of data for a given unit. It then sends an email with the result. MozReview-Commit-ID: 5muzykWkjEg
15fb4eddf3c68916132f0d74441d512894b75fc1: ansible/hg-ssh: add a systemd service for hg bundle generation
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 16:54:05 -0700 - rev 362591
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: add a systemd service for hg bundle generation This is part 1 of replacing the existing CRON with a systemd timer unit.
8c34717860687a8969f43117f76fb394ea60cb3d: ansible: add hgssh1 (bug 1267793)
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 15:09:28 -0700 - rev 362590
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible: add hgssh1 (bug 1267793) hgssh1 is back online. We can add it back to Ansible. We also add a host group containing hgssh3 to denote it as the master.
904566f0d9e5390e333526bfc530cd050fd4f6c4: ansible/hg-ssh: create v-c-t clone when it doesn't exist
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 14:41:19 -0700 - rev 362589
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
ansible/hg-ssh: create v-c-t clone when it doesn't exist This was preventing Ansible from working on a freshly provisioned server.
528b2f35323195ffa56a0b20ac38540759ca9da9: docs: update pushlog docs to reference change notification services
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 11:44:46 -0700 - rev 362588
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
docs: update pushlog docs to reference change notification services
30f52614e6af70e7829f1f6e93c12b19bc99b82d: docs: fix bad link
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 11:44:35 -0700 - rev 362587
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
docs: fix bad link
c96ce58178da5c894297ea30f818d9c2918a51a9: vcsreplicator: check whether local repository exists (bug 1267619)
Gregory Szorc <gps@mozilla.com> - Tue, 26 Apr 2016 08:50:18 -0700 - rev 362586
Push 16998 by rwood@mozilla.com at Mon, 02 May 2016 19:42:03 +0000
vcsreplicator: check whether local repository exists (bug 1267619) If the local repository doesn't exist, we should skip sending a notification. This should be rare. But it can happen when e.g. user repos are created and deleted in rapid order.
(0) -300000 -100000 -30000 -10000 -3000 -1000 -300 -100 -50 -20 +20 +50 +100 +300 +1000 +3000 +10000 +30000 +100000 +300000 tip