proxmox-mirrors/mirror_frr
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_frr synced 2025-08-13 22:26:14 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

lib: Figure out if we are being starved for cpu

Browse Source 
If a thread timer should have popped CPU_CONSUMED_CHECK
seconds in the past, and we are only handling it now.  Consider
the thread starved and notice it.

Signed-off-by: Donald Sharp <sharpd@nvidia.com>
This commit is contained in:
Donald Sharp 2021-02-03 09:13:59 -05:00
parent b542c4a3c7
commit ab01a00176
3 changed files with 26 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
lib/lib_errors.c
View File

@ -56,6 +56,12 @@ static struct log_ref ferr_lib_warn[] = {
.description = "The Event subsystem has detected a slow process, this typically indicates that FRR is having trouble completing work in a timely manner. This can be either a misconfiguration, bug or some combination thereof. In this case total WALL time was over 5 seconds. Which indicates that FRR might be having trouble being scheduled or some system call is delaying",
.suggestion = "Gather log data and open an Issue",
},
{
.code = EC_LIB_STARVE_THREAD,
.title = "The Event subsystem has detected a thread starvation issue",
.description = "The event subsystem has detected a thread starvation issue. This typically indicates that the system FRR is running on is heavily loaded and this load might be impacting FRR's ability to handle events in a timely fashion",
.suggestion = "Gather log data and open an Issue",
},
{
.code = EC_LIB_NO_THREAD,
.title = "The Event subsystem has detected an internal FD problem",

1
lib/lib_errors.h
View File

@ -46,6 +46,7 @@ enum lib_log_refs {
EC_LIB_LINUX_NS,
EC_LIB_SLOW_THREAD_CPU,
EC_LIB_SLOW_THREAD_WALL,
EC_LIB_STARVE_THREAD,
EC_LIB_NO_THREAD,
EC_LIB_RMAP_RECURSION_LIMIT,
EC_LIB_BACKUP_CONFIG,

19
lib/thread.c
View File

@ -1651,12 +1651,31 @@ static void thread_process_io(struct thread_master *m, unsigned int num)
static unsigned int thread_process_timers(struct thread_master *m,
struct timeval *timenow)
{
struct timeval prev = *timenow;
bool displayed = false;
struct thread *thread;
unsigned int ready = 0;
while ((thread = thread_timer_list_first(&m->timer))) {
if (timercmp(timenow, &thread->u.sands, <))
break;
prev = thread->u.sands;
prev.tv_sec += 4;
/*
* If the timer would have popped 4 seconds in the
* past then we are in a situation where we are
* really getting behind on handling of events.
* Let's log it and do the right thing with it.
*/
if (timercmp(timenow, &prev, >)) {
if (!displayed)
flog_warn(
EC_LIB_STARVE_THREAD,
"Thread Starvation: %pTHD was scheduled to pop greater than 4s ago",
thread);
displayed = true;
}
thread_timer_list_pop(&m->timer);
thread->type = THREAD_READY;
thread_list_add_tail(&m->ready, thread);