On 10/21/19 6:00 PM, Alan Brown wrote: > On 21/10/2019 16:13, Dimitri Maziuk via Bacula-users wrote: >> On 10/21/2019 9:14 AM, Clark, Patti via Bacula-users wrote: >> ... >>> My question (rhetorical), why do you want to use an OS that is >>> recently released and still bleeding for a backup server? I >>> understand upgrade headaches, but backup servers need to be rock solid. >> Why use any RHEL > 6 for any kind of server? >> > Because RHEL6 is EOL and won't even install on newer hardware? > > (EOS is some time off, but it's on life support mode now - bug/security > fixes only and one of the more annoying "features" is that RHEL6 doesn't > support ECN properly - it has ECN but not ECN fallback, so it can hammer > the living daylights out of routers. In addition it's quite difficult to > fully secure a php-using webserver on RHEL6 using stock RPMs) > > > As for "Why use RHEL?" - Corporate policy. The scientific Linux > community standardised on redhat 20 years ago and they're unlikely to > change anytime soon despite arguments about other flavours being > "better" (Most of them are still using Fortran code rooted in f77 which > breaks if you attempt to update it, or IDL despite it being "simply awful") > > > I spend an awful lot of time making things work in a RHEL environment > when it "just works" in Debianish ones, but the choice of environment > isn't my call. > > Everything has its positives and negatives - as Kern knows, I _really_ > want Bacula to support the IBM lin_tape driver because it works far > better with LTOs and changers - particularly in FC/SAS multipath > environments but Baculasystems won't do that because the st driver works > just fine(*) and the RHEL vs Debian arguments work in much the same way.
My comments for the above paragraph: Yes, a lot of decisions in Bacula and elsewhere are tradeoffs. In the case of the IBM lin_tape driver, I am sure that what you write about it being better is correct. However there needs to be a technical person familiar with Bacula to implement it, and that person needs access to such a tape setup (probably onsite). I am past the time when I am available to develop major new "tricky" technology, so this would leave either a community submission, or Bacula Systems to implement it. It is quite possible that Bacula Systems would implement this feature if it turns out to be something that a number (3 or more) of important customers want -- i.e. it is mainly a cost/benefit question for them. Best regards, Kern > > > (*) For some values of "fine" - st and sg don't handle multipathing _at > all_ and you end up with door locking contention issues that have caused > us issues for 15 years > > > > _______________________________________________ > Bacula-users mailing list > Bacula-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/bacula-users _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users