Upon @faidon's suggestion, we should configure netconsole on at least upload@esams to log kernel messages to a central host, to ensure that we don't miss kernel messages potentially emitted at crash time but before the hosts manage to write them to disk.
Related Gerrit Patches:
|operations/puppet : production||netconsole: add server to esams ganeti|
|operations/puppet : production||netconsole: add netconsole::server|
|operations/puppet : production||netconsole:: rename to netconsole::client|
|operations/puppet : production||cache: enable netconsole on all upload@esams|
|operations/puppet : production||profile::netconsole: use ipaddress and interface_primary|
|operations/puppet : production||cache: configure netconsole on cp3061|
In case it is helpful: we can reuse the centrallog hosts in codfw/eqiad. For site-local netconsole instead we'll need to setup local syslog collectors anyways (on ganeti VMs) for network devices syslog, so we could piggyback on those.