- Split-brain recovery in headless enable, IMMND may expected restart.
If AMFND not wait IMMND restart but reinit CLM, CLM callback trigger,
clm_to_amf_node() is called then AMFND stuck in init IMM OM causes delay
restart IMMND, delay resend node_up then AMFD will order reboot node.
- Slow down this reinit CLM help avoid node unexpected reboot order.
---
src/amf/amfnd/avnd_cb.h | 1 +
src/amf/amfnd/clc.cc | 7 +++++++
src/amf/amfnd/clm.cc | 9 +++++++++
src/amf/amfnd/main.cc | 1 +
4 files changed, 18 insertions(+)
diff --git a/src/amf/amfnd/avnd_cb.h b/src/amf/amfnd/avnd_cb.h
index 8b0cc2304..0fa0590ff 100644
--- a/src/amf/amfnd/avnd_cb.h
+++ b/src/amf/amfnd/avnd_cb.h
@@ -125,6 +125,7 @@ typedef struct avnd_cb_tag {
SaTimeT scs_absence_max_duration;
/* the timer for supervision of the absence of SC */
AVND_TMR sc_absence_tmr;
+ bool immnd_down;
} AVND_CB;
#define AVND_CB_NULL ((AVND_CB *)0)
diff --git a/src/amf/amfnd/clc.cc b/src/amf/amfnd/clc.cc
index f78e1a707..e8e309051 100644
--- a/src/amf/amfnd/clc.cc
+++ b/src/amf/amfnd/clc.cc
@@ -3333,6 +3333,13 @@ uint32_t avnd_comp_clc_cmd_execute(AVND_CB *cb,
AVND_COMP *comp,
// outcome of command is reported in comp_clc_resp_callback()
}
+ if (comp->name.find("safComp=IMMND,") != std::string::npos) {
+ if (cmd_type == AVND_COMP_CLC_CMD_TYPE_CLEANUP)
+ cb->immnd_down = true;
+ else if (cmd_type == AVND_COMP_CLC_CMD_TYPE_INSTANTIATE)
+ cb->immnd_down = false;
+ }
+
TRACE_2("success");
goto done;
diff --git a/src/amf/amfnd/clm.cc b/src/amf/amfnd/clm.cc
index 06eb229c7..01151b2a7 100644
--- a/src/amf/amfnd/clm.cc
+++ b/src/amf/amfnd/clm.cc
@@ -312,6 +312,15 @@ done:
static void *avnd_clm_init_thread(void *arg) {
TRACE_ENTER();
AVND_CB *cb = static_cast<AVND_CB *>(arg);
+ int count = 0;
+
+ while (cb->immnd_down && count++ < 100)
+ osaf_nanosleep(&kHundredMilliseconds);
+ /* Split-brain recovery in headless enable, IMMND may expected restart.
+ * If AMFND not wait IMMND restart but reinit CLM, CLM callback trigger,
+ * clm_to_amf_node() is called then AMFND stuck in init IMM OM causes delay
+ * restart IMMND, delay resend node_up then AMFD will order reboot node.
+ * Slow down this reinit CLM help avoid node unexpected reboot order */
avnd_clm_init(cb);
diff --git a/src/amf/amfnd/main.cc b/src/amf/amfnd/main.cc
index d7857fabe..cef0543e6 100644
--- a/src/amf/amfnd/main.cc
+++ b/src/amf/amfnd/main.cc
@@ -334,6 +334,7 @@ AVND_CB *avnd_cb_create() {
cb->is_avd_down = true;
cb->amfd_sync_required = false;
+ cb->immnd_down = false;
// retrieve hydra configuration from IMM
hydra_config_get(cb);
--
2.17.1
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel