Freeradius 3.0.17 seems to have deadlocked on me.

[Alan DeKok]

On Jul 19, 2019, at 2:40 PM, Matthew Schumacher <matt.s at aptalaska.net> wrote:
> I upgraded to 3.0.19 and just had the same issue again:

  Dang...

> Wed Jul 17 00:35:30 2019 : Error: Received conflicting packet from client 64.186.123.90 port 46124 - ID: 37 due to unfinished request in module sql_wifi.  Giving up on old request.

  That's blocked, but there aren't too many of those messages.

> Wed Jul 17 00:35:33 2019 : Error: Received conflicting packet from client 64.186.123.90 port 46124 - ID: 37 due to unfinished request in module sql_wifi.  Giving up on old request.
> Wed Jul 17 00:35:38 2019 : Error: Received conflicting packet from client 64.186.123.40 port 42193 - ID: 184 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:39 2019 : Error: Received conflicting packet from client 64.186.122.9 port 39175 - ID: 127 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:39 2019 : Error: Received conflicting packet from client 64.186.122.25 port 45721 - ID: 183 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:40 2019 : Error: Received conflicting packet from client 64.186.122.14 port 37015 - ID: 95 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:41 2019 : Error: Received conflicting packet from client 64.186.123.62 port 56187 - ID: 202 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:41 2019 : Error: Received conflicting packet from client 64.186.123.40 port 42193 - ID: 184 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:42 2019 : Error: Received conflicting packet from client 64.186.122.9 port 39175 - ID: 127 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:42 2019 : Error: Received conflicting packet from client 64.186.123.57 port 59432 - ID: 59 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:42 2019 : Error: Received conflicting packet from client 64.186.122.25 port 45721 - ID: 183 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:43 2019 : Error: Received conflicting packet from client 64.186.122.14 port 37015 - ID: 95 due to unfinished request in module <queue>.  Giving up on old request.
> Wed Jul 17 00:35:43 2019 : Error: (685424) Ignoring duplicate packet from client 64.186.96.236 port 44092 - ID: 208 due to unfinished request in component <core> module <queue>
> Wed Jul 17 00:35:44 2019 : Error: Received conflicting packet from client 64.186.123.62 port 56187 - ID: 202 due to unfinished request in module <queue>.  Giving up on old request.

  Yeah... once all of the worker threads are blocked, the input queue gets blocked, too.

> It looks like my sql_wifi module is too slow and needs to be better, but it shouldn't deadlock the entire server unto the point of needing to be killed -9.

  Well, if the SQL query never returns, it will get blocked.  And we don't really control what the SQL client libraries do.  Some are *entirely* blocking.

> Anything I can provide the devs that would help identify where the server is locking up?  Backtraces?

  Find out why the SQL module is blocking.  Even if the SQL back-end takes time to respond, it should *eventually* make progress.

  If this is for PostgreSQL, then try using the v3.0.x branch.  Terry Burton put some patches in which make it better respect the query_timeout parameter.  That may help.

  Alan DeKok.