picnoir
/
Systemd
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

core: change default trigger limits for socket units 

Let's lower the default values a bit, and pick different defaults for
Accept=yes and Accept=no sockets.

Fixes: #3167
This commit is contained in:
Lennart Poettering 2016-05-05 13:39:31 +02:00
parent 37818090c9
commit 1f15ce2846
2 changed files with 28 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
man/systemd.socket.xml
View File

@ -814,13 +814,14 @@
<listitem><para>Configures a limit on how often this socket unit my be activated within a specific time
interval. The <varname>TriggerLimitIntervalSec=</varname> may be used to configure the length of the time
interval in the usual time units <literal>us</literal>, <literal>ms</literal>, <literal>s</literal>,
<literal>min</literal>, <literal>h</literal>, … and defaults to 5s (See
<literal>min</literal>, <literal>h</literal>, … and defaults to 2s (See
<citerefentry><refentrytitle>systemd.time</refentrytitle><manvolnum>7</manvolnum></citerefentry> for details on
the various time units available). The <varname>TriggerLimitBurst=</varname> setting takes an integer value and
specifies the numer of permitted activations per time interval, and defaults to 2500 (thus by default
permitting 2500 activations per 5s). Set either to 0 to disable any form of trigger rate limiting. If the limit
is hit, the socket unit is placed into a failure mode, and will not be connectible anymore until
restarted. Note that this limit is enforced before the service activation is enqueued.</para></listitem>
the various time units understood). The <varname>TriggerLimitBurst=</varname> setting takes a positive integer
value and specifies the number of permitted activations per time interval, and defaults to 200 for
<varname>Accept=yes</varname> sockets (thus by default permitting 200 activations per 2s), and 20 otherwise (20
activations per 2s). Set either to 0 to disable any form of trigger rate limiting. If the limit is hit, the
socket unit is placed into a failure mode, and will not be connectible anymore until restarted. Note that this
limit is enforced before the service activation is enqueued.</para></listitem>
</varlistentry>
</variablelist>

22
src/core/socket.c
View File

@ -100,7 +100,8 @@ static void socket_init(Unit *u) {
s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID;
RATELIMIT_INIT(s->trigger_limit, 5*USEC_PER_SEC, 2500);
s->trigger_limit.interval = USEC_INFINITY;
s->trigger_limit.burst = (unsigned) -1;
}
static void socket_unwatch_control_pid(Socket *s) {
@ -328,6 +329,25 @@ static int socket_add_extras(Socket *s) {
assert(s);
/* Pick defaults for the trigger limit, if nothing was explicitly configured. We pick a relatively high limit
* in Accept=yes mode, and a lower limit for Accept=no. Reason: in Accept=yes mode we are invoking accept()
* ourselves before the trigger limit can hit, thus incoming connections are taken off the socket queue quickly
* and reliably. This is different for Accept=no, where the spawned service has to take the incoming traffic
* off the queues, which it might not necessarily do. Moreover, while Accept=no services are supposed to
* process whatever is queued in one go, and thus should normally never have to be started frequently. This is
* different for Accept=yes where each connection is processed by a new service instance, and thus frequent
* service starts are typical. */
if (s->trigger_limit.interval == USEC_INFINITY)
s->trigger_limit.interval = 2 * USEC_PER_SEC;
if (s->trigger_limit.burst == (unsigned) -1) {
if (s->accept)
s->trigger_limit.burst = 200;
else
s->trigger_limit.burst = 20;
}
if (have_non_accept_socket(s)) {
if (!UNIT_DEREF(s->service)) {