Legacy Equinix Metal data facility closures on November 30th, 2022

[richardlau]

We've been informed that Equinix are closing their legacy data facilities on November 30th, 2022. This will affect all of the instances we currently have hosted there and we'll need to move to "go-forward Equinix-based environments".

Currently we're spread over two projects -- Node.js and Works on ARM. The Linux ARM64 release machines are older ARM64 hardware (c2.large.arm) that are only in the legacy data facilities and are most at risk.

Node.js

host	type	desc	status
release-equinix-ubuntu2004-docker-arm64-1	c2.large.arm	Linux ARM64 release machine	migrated to OSUOSL
release-equinix-ubuntu2004-docker-arm64-2	c2.large.arm	Linux ARM64 release machine	migrated to OSUOSL

Works on ARM:

host	type	desc	status
test-equinix-ubuntu2004-arm64-1	x.large.arm	Linux ARM64 test machine	Migrated
test-equinix-ubuntu2004-arm64-3	x.large.arm	Linux ARM64 test machine	Migrated
test-packetnet-ubuntu1804-x64-1	c1.small.x86	jenkins-workspace-4	Migrated to Node.js project at Equinix
test-packetnet-ubuntu1804-x64-2	c1.small.x86	jenkins-workspace-5	Migrated to Node.js project at Equinix

Actions

• Contact Equinix to discuss migration plan for the ARM hardware -- Email has been sent to @vielmetti
• Migrate ARM64 release machines -- ansible: add new OSUOSL arm64 release machines #3051
• Migrate ARM64 test machines
• Migrate the two Jenkins workspace machines from Works on ARM into Node.js -- ansible: add new jenkins-workspace machines #3036

[vielmetti]

Thanks @richardlau . Here's what I know so far.

The Jenkins machines should be the easiest to move of this bunch. We have a "c3.small" machine type which is of comparable size which should do the trick.

For the x.large.arm systems, the team is working on a migration process that will likely result in us doing physical machine migration on your behalf, moving the systems from DFW2 to one of the DA data centers. Some of those details are TBD, so as of right now I an awaiting instructions. In an ideal situation you would see a little bit of downtime for the physical move, but no need to reconfigure.

For the two release machines (c2.large.arm) my recommendation at the moment is to migrate those services to another provider. We are decommissioning the Ampere eMag hardware these live on, and those machines are not being moved to new facilities. Our newer c3.large.arm are very similar to the x.large.arm systems you currently are using through Works on Arm - large bare metal systems with high core counts - which makes them a poor fit to be deployed as release-only machines.

Happy to discuss any of this further (here, or in email, or by phone or conference call).

[mhdawson]

@vielmetti I think our challenge in terms of the suggestion for the release machines is that we don't have other providers with arm machines. One alternative is to try to use one larger machine and docker for the two release machines. Would it be possible to get one to do that?

[vielmetti]

@mhdawson In the time since we started this effort (at Packet), there have been a number of vendors that have announced and released arm64 support, many of which have done it with VMs that allow for the provider to offer up a smaller slice of a machine. AWS Graviton has been in the market for a while, and Google and Microsoft have made announcements.

Arm does have a program to enable access to these smaller systems for developers, an extension of the Works on Arm program described here

https://www.arm.com/solutions/infrastructure/works-on-arm

Please take a look around for some alternatives - as I say there are a lot more choices than there were just a few years ago! - some one of those may be economical or free to suit your needs.

[mhdawson]

@vielmetti ok we'll try to explore other options.

[richardlau]

We have replacement arm64 release machines at OSUOSL now (#3051).

The remaining outstanding item here is the Altras (x.large.arm) in the test CI that I guess we're waiting to here further information on from Equinix.

[vielmetti]

Thanks @richardlau . When the replacement arm64 systems are 100% a-ok, can you go into the Equinix portal and delete/destroy the old systems that are no longer in service?

[richardlau]

Thanks @richardlau . When the replacement arm64 systems are 100% a-ok, can you go into the Equinix portal and delete/destroy the old systems that are no longer in service?

@vielmetti I've deleted the old systems that have already been migrated (the two c2.large.arm's and the c1.small.x86's).

The remaining servers affected by the data facility closures are the two Altras (x.large.arm). Do we need to initiate something to get these machines moved to a data facility that is not being closed or will we be contacted with further details?