Thank you for the reply.

We are using symas rpm pkgs.

Of course, test env first but wanted to run it here , before i start

Thank you On Jan 31, 2022, 2:08 PM -0500, Shawn McKinney smckinney@symas.com, wrote:

...

On Jan 30, 2022, at 2:52 PM, Dave Macias davama@gmail.com wrote:

3-way MMR (syncrepl) with centos 7 using v2.6

We want to add a 4th master but with Rocky Linux 8.

Why rocky? because eventually we will migrate the 3 masters to rocky linux too. But that will take some time…

Any concerns with mixing linux flavor but using same openldap version? Dont do it?

I wouldn’t anticipate any problems mixing different distros, all using the same version.

Of course you’d have different setups for ancillary concerns like logging, core dumps, etc.

Also, depending on what OpenLDAP packages you're using, e.g. built from source, vanilla Debian, LTB, Symas, the artifacts might well be placed in different locations and the systemd bits will vary.

But as usual, give it a try in a test env and see how it goes.

— Shawn

...

Any input is appreciated.

On Jan 31, 2022, at 1:25 PM, Shawn McKinney smckinney@symas.com wrote:

...

On Jan 31, 2022, at 1:16 PM, Dave Macias davama@gmail.com wrote:

Thank you for the reply.

We are using symas rpm pkgs.

Of course, test env first but wanted to run it here , before i start

Speaking of tests…

Here are the results of a test I ran today. 4 providers, each on their own distro:

slapd 1. Centos7 el7 slapd 2. AlmaLinux8 el8 slapd 3. Ubuntu20.04 focal slapd 4. Debian10 buster

All using using 2.6.1-2 packages, from our testing repo.

Pumped several hundred thousand entries into the cluster, via the openldap load balancer. As expected, no problems found.

The machines all behave the same and don’t care what operating system the others run.

Now, we could move into a convo about mixing and matching versions, i.e. 2.4. 2.5 and 2.6. Again, 'should work’, assuming the 2.4 is up-to-date, and they’re not configured ‘differently'. But, that’s another post, for another time :-)

— Shawn

In that case, there should be no variance in where the bits are placed.

Core dump enablement has changed between RHEL7 and 8 distros, but you probably already know that.

--On Wednesday, February 2, 2022 4:47 PM -0500 Dave Macias davama@gmail.com wrote:

This is awesome! Thank you Shawn!

Here are the results of a test I ran today. 4 providers, each on their own distro:

slapd 1. Centos7 el7 slapd 2. AlmaLinux8 el8 slapd 3. Ubuntu20.04 focal slapd 4. Debian10 buster

All using using 2.6.1-2 packages, from our testing repo.

Pumped several hundred thousand entries into the cluster, via the openldap load balancer.  As expected, no problems found.

The machines all behave the same and don't care what operating system the others run.

Which is generally what I'd expect, given LDAP is a protocol. The areas where there could be issues due to distributions would be things like:

a) Issues that exist in the kernel that affect things like threading performance b) Filesystem differences that affect I/O rates c) sysctl tuning differences for the TCP layer

One area that could cause issues if using cn=config replication and OpenLDAP binaries that are linked to the system SSL libraries is the differences between GnuTLS and OpenSSL, as the configurations could be incompatible. But in the more general case, the underlying distribution should be immaterial, especially if using something like the packages from Symas which are consistent in regards to OpenSSL and other support libraries.

Regards, Quanah