Apply recovery pause before writes, after commits, and do not apply it to syncs to not block EC pools from functioning

test-fix-ec-unknown-state-51
Vitaliy Filippov 2024-02-11 12:19:08 +03:00
parent bb2f395f1e
commit 978bdc128a
3 changed files with 24 additions and 3 deletions

View File

@ -283,6 +283,7 @@ class osd_t
void exec_sync_stab_all(osd_op_t *cur_op);
void exec_show_config(osd_op_t *cur_op);
void exec_secondary(osd_op_t *cur_op);
void exec_secondary_real(osd_op_t *cur_op);
void secondary_op_callback(osd_op_t *cur_op);
// primary ops

View File

@ -442,7 +442,7 @@ void osd_t::tune_recovery()
if (recovery_target_sleep_count < recovery_tune_agg_interval)
recovery_target_sleep_count++;
recovery_target_sleep_us = recovery_target_sleep_total / recovery_target_sleep_count;
if (log_level > 4)
if (log_level > 1)
{
printf(
"[OSD %lu] auto-tune: client util: %.2f, recovery util: %.2f, lat: %lu us -> target util %.2f, delay %lu us\n",

View File

@ -42,8 +42,10 @@ void osd_t::secondary_op_callback(osd_op_t *op)
int retval = op->bs_op->retval;
delete op->bs_op;
op->bs_op = NULL;
if (op->is_recovery_related() && recovery_target_sleep_us)
if (op->is_recovery_related() && recovery_target_sleep_us &&
op->req.hdr.opcode == OSD_OP_SEC_STABILIZE)
{
// Apply pause AFTER commit. Do not apply pause to SYNC at all
if (!op->tv_end.tv_sec)
{
clock_gettime(CLOCK_REALTIME, &op->tv_end);
@ -59,7 +61,25 @@ void osd_t::secondary_op_callback(osd_op_t *op)
}
}
void osd_t::exec_secondary(osd_op_t *cur_op)
void osd_t::exec_secondary(osd_op_t *op)
{
if (op->is_recovery_related() && recovery_target_sleep_us &&
op->req.hdr.opcode != OSD_OP_SEC_STABILIZE && op->req.hdr.opcode != OSD_OP_SEC_SYNC)
{
// Apply pause BEFORE write/delete
tfd->set_timer_us(recovery_target_sleep_us, false, [this, op](int timer_id)
{
clock_gettime(CLOCK_REALTIME, &op->tv_begin);
exec_secondary_real(op);
});
}
else
{
exec_secondary_real(op);
}
}
void osd_t::exec_secondary_real(osd_op_t *cur_op)
{
if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
{