this makes it much easier to reason about the initialization.
for production, init scripts should enforce the dependency order
for testing, provide your own if desired
//cc MagnusS
also move enable_stats and retry_connections up (before exit codes)
* Make albatrossd wait for other sockets to appear if --retry-connections is specified
* Connect to log/console before creating main socket
* Disable stats by default unless --enable stats is specified
regression introduced in dfd22be62b
(which prepared dumping to a file, and restart-on-failure)
this is not desired from a bandwidth point of view, if it is
considered to be a useful feature, there should be a separate
command for it
- console creates the run directories (albatross albatross/fifo albatross/util)
-> log depends on console
- albatross/fifo needs special permission (albatross_daemon creates fifo there, albatross_console needs to read them)
- use /var/lib/albatross/albatross.log for the log file
- only require root in albatross_daemon, use user albatross in albatross_console and albatross_log
- defer albatross_stat for now
it used to only contain service, and used the same string for the bridge. This
is not flexible enough to run off-the-shelf unikernels (configured for bridge
"service" and "management" on multi-homed servers). The old behaviour is the
new default (i.e. "--net=service" creates and attaches a tap device to bridge
"service", and passes "--net:service=tapYY" to the solo5 tender). But it is more
flexible now: "--net=service:other-bridge" will create a tap device attached to
"other-bridge" and pass "--net:service=tapYY" to the tender. This way, there's
no need to match bridge names on the actual server with network device names of
the unikernels.
NB: this is (mostly) backwards-compatible: the on-disk data structures are
versioned (and the version is bumped with this PR), an old albatross client can
send "create" commands to a new server. But a new client will get a parse error
from an old server - which is fine taking into consideration the deployment
base.
the socket lock was removed in 58c34907 in favour of a global lock -- but the
bootup used an iter_p for create (which is usually called with the lock being
held by the caller)
by Lwt_unix.wait_read (in read_console, called asynchronously in a fresh task)
and only then convert to a Lwt_io.t (of_fd ~mode:Input) -- thanks to @cfcs
This allows albatross to be run on Linux, where the former code lead to:
- open FIFO
- read <- EOF (although no writer was connected yet)
this works nicely around ocsigen/lwt#741
