Hi Anders ,
On 8/15/2017 6:13 PM, Anders Widell wrote:
Is it a pure performance optimization, or is there some potential
problem that can happen if two nodes discover each other at the same
time and estabilsh TCP connections in both directions? I am seeing
error log messages that
Hi Zoran,
ACK, not tested.
-AVM
On 8/7/2017 8:33 PM, Zoran Milinkovic wrote:
CLM selection object is initially set to -1. Included CLM selection
object in poll will be ignored until CLM selection object is created
and set to fds[FD_CLM].
---
src/imm/immnd/immnd_main.c | 4 +---
1 file
Hi Zoran ,
ACK , with following , not tested.
Please check that the `immnd_cb->clm_hdl` cleared on Headless case.
-AVM
On 8/9/2017 5:14 PM, Zoran Milinkovic wrote:
Initially CLM handle is set to 0. When CLM handle is initialized,
it can be initialized again only when saClmDispatch returns
Ack with comments 1) and 2) below. In addition, I think we should leave
this new feature disabled by default until we have investigated the
possible stability problems. I have only seen the core dump once, and I
don't think it is caused directly by your patch but rather it is
something in the
ack, code review only. One minor comment below.
/Regards HansN
On 07/05/2017 10:45 AM, Gary Lee wrote:
---
src/amf/amfnd/avnd_cb.h | 3 +-
src/amf/amfnd/avnd_di.h | 36 --
src/amf/amfnd/avnd_mds.h | 11 +-
src/amf/amfnd/di.cc | 321
A question to Mahesh: what is the reason behind the following few lines
of code:
if (initial_discovery_phase == true) {
if (node.node_id < dtms_cb->node_id) {
TRACE(
"DTM: received node_id is less than local node_id
dropping message cluster_id: %d
Yes I agree, I think it is more flexible to have a new configuration
option for this.
Some comments on the patch:
1) The new configuration option should specify broadcast interval time
in the unit of milliseconds rather than seconds, to make it possible to
configure sub-second intervals.
