Overpass API developpement

[Roland Olbricht <>]

Hello,

a short remark: It should never happen as part of regular operations 
that you need to delete any lock files. Leftover lock files are always 
an indicator of a unexpectedly terminated process.

> I tried to simulate this via vagrant suspend + up and found that
> osm3s_query immediately returned with:
>
> runtime error: open64: 32 Broken pipe /osm3s_v0.7.54_osm_base
> Dispatcher_Client::request_read_and_idx::socket::1
>
> ... while dispatcher crashed with a SIGBUS error right after it received
> a command from osm3s_query.

Thank you for the simulation. Accroding to the information I find about 
SIGBUS, this happens if a shared memory file has been deleted. I found 
no information that it is related to sockets.

Can you check whether both the shared memory and the socket exist before 
the incident? I.e. please run the commands

ls -l $DB_DIR/osm*

ls -l /dev/shm/ | grep osm

ps -ef | grep dispatcher

This should give us an idea of the cause of events.

At mmd: what happens if you comment out line template_db/dispatcher.cc:768
      *(uint32*)dispatcher_shm_ptr = 0;

This should give a clue about the order of events (whether the socket or 
the shared memory fails first). A broken pipe for a client if the server 
crashes would be expected behaviour. But this does no explain why the 
dispatcher has crashed.

BTW: thank you for using the mailing list. This makes answering from the 
train a lot easier.

Best regards,

Roland