The audit of repos identified ~13000 corrupted rpms
So far ~5000 have been restored from redundant copies within cengn.
Aprox 7500 have yet to be assessed. Those remaining rpms are inputs or outputs of actual builds. The rpms produced by builds might carry the same files name, but may contain different content due to embedded file dates if not real changes. Extra care is required to restore from the correct source. If an exact match is unavailable, I may be forced to use a source that is 'near' in time and therefore less likely to contain anything other than trivial file timestamp deltas. I'm still pondering the implications of such a substitution.
That leaves ~500 having no backup within cengn. Upstream availability has yet to be determined.
Scott
On 2021-01-26 1:28 a.m., Scott Little wrote:
The jenkins in the cengn build is offline while we repair the file system on the mirror. The build machine both read from, and writes to the mirror, and I don't want it to cause further corruptions while the mirror is being repaired.
The cengn mirror has been cut over to the new storage back end. However any corrupt files from the old storage back end will remain corrupted on the new one.
Restoration of rpm repos will proceed in several phases:
1) An audit of all rpm repos is underway. Rpm's that fail their checksum are being removed and repodata will be updated to reflect the missing RPMS.
2) The mirror update job (downloading rpms from upstream rpm repos) will be run to restore as many missing files as possible. Many rpms may no longer be available as we kept old versions of packages that upstream drops when a new version is available,
3) If rpms that are not available from upstream, but are required for our build, we can attempt to restore them from one of several places. e.g. the inputs directory of various builds, or the downloads directory on the cengn build server (independent storage from the mirror)
4) Resume building new loads
All of the release iso's on cengn were corrupt.
I've been able to restore 4.0.1 and 2.0.1 from the original build artifacts.
Original build artifacts for 3.0.0 are missing. I do have build artifacts for builds that were intended to lead up to a 3.0.1 release. If a verifiable copy of 3.0.0 can't be found, we may want to consider issuing a 3.0.1. Alternatively I can rebuild 3.0.0, but the file checksums wont match the original build due to embedded timestamps.
Scott
On 2021-01-25 11:08 a.m., Little, Scott wrote:
Work continues at CENGN repair damaged repos and to move us to a new backend infrastructure with better redundancy.
I still recommend that you avoid using download_mirrir.sh until the repos are fully repaired. I don’t have a firm ETA for that, perhaps two before we have full confidence in the repairs.
Scott
From: Scott Little <scott.little@windriver.com>
Sent: January 23, 2021 2:24 AM
To: starlingx-discuss@lists.starlingx.io <starlingx-discuss@lists.starlingx.io>
Subject: Re: [Starlingx-discuss] Cengn Mirror server in down?!CENGN is up, but there seems to be significant damage to the repositories.
We are investigate options to restore the repositories to health.
I do NOT recommend running download_mirror.sh at this time.
Scott
On 2021-01-22 3:53 p.m., Scott Little wrote:
CENGN remains out of service.
A new problem has been identified with ceph storage.
They continue to work the outage.
Scott
On 2021-01-22 9:55 a.m., Scott Little wrote:
CENGN remains out of service
AC to the server room has been restored. They are now working some firewall issues.
They hope to have it restored in a few hours.
Scott
On 2021-01-21 10:07 a.m., Panech, Davlet wrote:
The hosting company had an issue with their servers, they are working on it. They hope to be back online mid afternoon (Jan 21, Eastern Standard Time).
From: Panech, Davlet <Davlet.Panech@windriver.com>
Sent: January 21, 2021 9:56 AM
To: Dimofte, Alexandru <alexandru.dimofte@intel.com>; starlingx-discuss@lists.starlingx.io <starlingx-discuss@lists.starlingx.io>
Subject: Re: Cengn Mirror server in down?!Thanks Alexandru, we are looking into it.
From: Dimofte, Alexandru <alexandru.dimofte@intel.com>
Sent: January 21, 2021 3:24 AM
To: starlingx-discuss@lists.starlingx.io <starlingx-discuss@lists.starlingx.io>
Subject: [Starlingx-discuss] Cengn Mirror server in down?!Hello guys,
I observed, trying to connect to http://mirror.starlingx.cengn.ca/mirror/starlingx/master/centos/ , that cengn mirror server is down.
Best regards,
Alex
Dimofte Alexandru
Software Engineer
STARLINGX TEAM
Skype no: +40 336403734
Personal Mobile: +40 743167456
Intel Romania
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss