From denis at ovsienko.info Thu Apr 9 14:50:36 2026 From: denis at ovsienko.info (Denis Ovsienko) Date: Thu, 9 Apr 2026 13:50:36 +0100 Subject: cfarm210 is very slow Message-ID: <20260409135036.78a1054c@atlas-11> Hello. My first attempt to log into cfarm210 today failed after several minutes: $ ssh cfarm210 Last login: Sat Jan 17 14:03:14 2026 from 129.151.92.78 Connection to cfarm210.cfarm.net closed. The second time I was watching "ssh -v", which also took several minutes, but eventually succeeded. From the rate of the messages it looked as if everything was consistently slow, rather than a particular step. Once logged in, the load average (1.52) didn't explain it, but disk access was unusually slow when I ran "autoconf --version", then "find /usr/local/" would print some lines quickly and stop for a while, then resume, then stop again and so on. So this could be either a failing disk, as was the case in August 2024 with cfarm211, or sshd trying to consume more entropy than the host can provide (if you temporarily point it to /dev/urandom instead of /dev/random and the SSH delays disappear, that's the case). Anyway, the thing I was looking for on cfarm210 in the first place is the Autoconf version, and it is 2.72. Would it be difficult to have the same or at least 2.71 in the following Buildbot builds? libpcap-solaris10-amd64 libpcap-solaris10-i386 libpcap-solaris10-sparc libpcap-solaris10-sparcv9 libpcap-solaris11-amd64 libpcap-solaris11-i386 libpcap-solaris11-sparc libpcap-solaris11-sparcv9 tcpdump-solaris10-amd64 tcpdump-solaris10-i386 tcpdump-solaris10-sparc tcpdump-solaris10-sparcv9 tcpdump-solaris11-amd64 tcpdump-solaris11-i386 tcpdump-solaris11-sparc tcpdump-solaris11-sparcv9 -- Denis Ovsienko From dam at opencsw.org Thu Apr 9 15:56:47 2026 From: dam at opencsw.org (Dagobert Michelsen) Date: Thu, 9 Apr 2026 15:56:47 +0200 Subject: cfarm210 is very slow In-Reply-To: <20260409135036.78a1054c@atlas-11> References: <20260409135036.78a1054c@atlas-11> Message-ID: <774F7FAE-6D56-45C9-A51C-1E5FC51453F8@opencsw.org> Hi Denis, > Am 09.04.2026 um 14:50 schrieb Denis Ovsienko via buildfarm : > My first attempt to log into cfarm210 today failed after several > minutes: > > $ ssh cfarm210 > Last login: Sat Jan 17 14:03:14 2026 from 129.151.92.78 > Connection to cfarm210.cfarm.net closed. > > The second time I was watching "ssh -v", which also took several > minutes, but eventually succeeded. From the rate of the messages it > looked as if everything was consistently slow, rather than a particular > step. > > Once logged in, the load average (1.52) didn't explain it, but disk > access was unusually slow when I ran "autoconf --version", then "find > /usr/local/" would print some lines quickly and stop for a while, then > resume, then stop again and so on. So this could be either a failing > disk, as was the case in August 2024 with cfarm211, or sshd trying to > consume more entropy than the host can provide (if you temporarily > point it to /dev/urandom instead of /dev/random and the SSH delays > disappear, that's the case). I killed quite a lot of hung sendmail processes in the zones on the machine and now everything seems reasonable fast again for me. > Anyway, the thing I was looking for on cfarm210 in the first place is > the Autoconf version, and it is 2.72. Would it be difficult to have the > same or at least 2.71 in the following Buildbot builds? The buildbot hosts run in fact on another machine. The GCC zones were updated to fractalkit in /usr/local. This is not feasable for the buildbot zones as these are the OpenCSW build zones which should have no other packages than the Oracle ones and OpenCSW. My update effort have somewhat stalled as there are some test failures for autoconf 2.73 on Solaris 10. I may give it another try in the near future, please feel free to ping me again. Best regards ? Dago -- "You don't become great by trying to be great, you become great by wanting to do something, and then doing it so hard that you become great in the process." - xkcd #896