# zuul1004.eqiad.wmnet - 2026-10-02 # Reimage failed (sre.hosts.reimage exit_code=99), install-console showed only a # BusyBox shell. The disks still had the software RAID (md0) and LVM (vg0) from # the previous install, so partitioning in the installer got stuck. # Same issue and fix as on zuul1005 on 2026-09-28. # # WARNING: this destroys all data on the host's disks. # connect to the installer shell (from a cumin host) sudo install-console zuul1004.eqiad.wmnet # 1. confirm the host and look at the disk layout hostname; cat /proc/partitions; lvm pvs # -> zuul1004, disks sda sdb sdc sdd with 3 partitions each, md0, VG vg0 on /dev/md0 # 2. deactivate and remove the LVM volume group lvm vgchange -an vg0; lvm vgremove -ff -y vg0 # -> Logical volume "swap", "root", "srv" successfully removed. Volume group "vg0" successfully removed # 3. remove the LVM label from the RAID device lvm pvremove -ff -y /dev/md0 # -> Labels on physical volume "/dev/md0" successfully wiped. # 4. stop the RAID array mdadm --stop /dev/md0 # -> mdadm: stopped /dev/md0 # 5. erase the RAID metadata from the member partitions for p in /dev/sd[a-d][1-3]; do echo "$p:"; mdadm --zero-superblock "$p"; done # -> "Unrecognised md component device" for sd[a-d]2 (not RAID members), expected # 6. wipe the partition tables (start of each disk, and the end where GPT keeps its backup) for d in sda sdb sdc sdd; do echo "wiping /dev/$d"; dd if=/dev/zero of=/dev/$d bs=1M count=16; sz=$(cat /sys/block/$d/size); dd if=/dev/zero of=/dev/$d bs=512 seek=$((sz - 32768)) count=32768; done; sync # 7. abort the waiting cookbook, then reimage again (from a cumin host) sudo cookbook sre.hosts.reimage --os trixie -t T427353 zuul1004