You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Teleport binaries that currently get packaged inside .rpm's that we publish to both rpm repo and those on our downloads page are currently built on Ubuntu that has a newer glibc and as such do not work on CentOS 7. I.e. these binaries are the same that are present inside our regular Linux 64-bit tarball instead of Linux 64-bit (RHEL/CentOS 7.x compatible).
These binaries require a newer version of glibc and as such do not run on CentOS 7:
[centos@ip-172-31-94-185 ~]$ wget https://get.gravitational.com/teleport-9.0.0-beta.1-1.x86_64.rpm
[centos@ip-172-31-94-185 ~]$ sudo rpm -i teleport-9.0.0-beta.1-1.x86_64.rpm
[centos@ip-172-31-94-185 ~]$ teleport version
teleport: /lib64/libc.so.6: version `GLIBC_2.25' not found (required by teleport)
teleport: /lib64/libc.so.6: version `GLIBC_2.18' not found (required by teleport)
Instead, our rpms should contain same binaries that are present inside CentOS 7 compatible package i.e. the ones that are built using CentOS 7 buildbox - the one that builds build-linux-amd64-centos7 and build-linux-amd64-centos7-fips targets.
For the previous major release, Teleport 8, this was addressed in a hacky way that only updated rpms published to the rpm repo to be CentOS 7 compatible and we never went back and addressed this properly. We should do that before Teleport 9 release. Here's the commit with the v8 fix for reference: c0a1e07.
The text was updated successfully, but these errors were encountered:
There's a few different solutions that I see here:
Start releasing Teleport in OS-specific channels. Lots of repos do this, it's just not something we've had a need for up to this point so I didn't mention it in my RFD.
Statically link our binaries. This would solve to issue permanently for all systems, but comes with the drawback of larger file sizes and needing to push an update for each glibc version.
Require clients to build Teleport themselves if they're on CentOS 7 or earlier. Not ideal as this raises the barrier for entry and increases maintenance.
Any thoughts on which solution would be best here?
Also, do we have a policy/list of what OS versions we support somewhere?
Description
Teleport binaries that currently get packaged inside .rpm's that we publish to both rpm repo and those on our downloads page are currently built on Ubuntu that has a newer glibc and as such do not work on CentOS 7. I.e. these binaries are the same that are present inside our regular
Linux 64-bit
tarball instead ofLinux 64-bit (RHEL/CentOS 7.x compatible)
.These binaries require a newer version of glibc and as such do not run on CentOS 7:
Instead, our rpms should contain same binaries that are present inside CentOS 7 compatible package i.e. the ones that are built using CentOS 7 buildbox - the one that builds
build-linux-amd64-centos7
andbuild-linux-amd64-centos7-fips
targets.For the previous major release, Teleport 8, this was addressed in a hacky way that only updated rpms published to the rpm repo to be CentOS 7 compatible and we never went back and addressed this properly. We should do that before Teleport 9 release. Here's the commit with the v8 fix for reference: c0a1e07.
The text was updated successfully, but these errors were encountered: