[tickets] [opensaf:tickets] #3097 mds: Incorrect fragment decoding for MDS_PROT_FCTRL_ID

2019-10-14 Thread Minh Hon Chau via Opensaf-tickets
- **status**: assigned --> fixed
- **assigned_to**: Minh Hon Chau -->  nobody 
- **Comment**:

commit 4779df8727c25af3006794d637bc2d3357bda50d
Author: Minh Chau 
Date:   Tue Oct 8 07:41:24 2019 +1100

mds: Enhance decoding for mds flow control message [#3097]




---

** [tickets:#3097] mds: Incorrect fragment decoding for MDS_PROT_FCTRL_ID**

**Status:** fixed
**Milestone:** 5.19.10
**Created:** Fri Oct 04, 2019 05:35 AM UTC by Minh Hon Chau
**Last Updated:** Mon Oct 07, 2019 12:14 AM UTC
**Owner:** nobody


The current code that decodes protocol identifier

~~~
  if ((pro_ver_ & MDS_PROT_VER_MASK) == MDS_PROT_FCTRL) {
// decode flow control sequence number
ptr = [HeaderMessage::FieldIndex::kFlowControlSequenceNumber];
fseq_ = ncs_decode_16bit();
// decode protocol identifier
ptr = [ChunkAck::FieldIndex::kProtocolIdentifier];
pro_id_ = ncs_decode_32bit();
if (pro_id_ == MDS_PROT_FCTRL_ID) {
  // decode message type
  ptr = [ChunkAck::FieldIndex::kFlowControlMessageType];
  msg_type_ = ncs_decode_8bit();
}
  }
~~~
  
In case of fragmentation, the last fragment has pro_ver = MDS_PROT_FCTRL, and 
coincidently the next 4 bytes (oct11 -> oct14)  has the same value to 
MDS_PROT_FCTRL_ID. Therefore, mds thinks this fragment as ack/nack message, 
which is wrong.


---

Sent from sourceforge.net because opensaf-tickets@lists.sourceforge.net is 
subscribed to https://sourceforge.net/p/opensaf/tickets/

To unsubscribe from further messages, a project admin can change settings at 
https://sourceforge.net/p/opensaf/admin/tickets/options.  Or, if this is a 
mailing list, you can unsubscribe from the mailing list.___
Opensaf-tickets mailing list
Opensaf-tickets@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets


[tickets] [opensaf:tickets] #3095 mds: Message loss after splitbrain with enabled MDS_TIPC_FCTRL_ENABLED

2019-10-14 Thread Minh Hon Chau via Opensaf-tickets
- **status**: review --> fixed
- **assigned_to**: Minh Hon Chau -->  nobody 
- **Comment**:

commit cafbc5d02c90b57c7c94a7735ce8e002224b3d6b
Author: Minh Chau 
Date:   Tue Oct 8 15:24:06 2019 +1100

mds: Improve error log for MDS_TIPC_FCTRL_ENABLED [#3095]

commit 13ccc878e7eac00c27cc66a9b353bb5a474b747f
Author: Minh Chau 
Date:   Tue Oct 8 15:24:06 2019 +1100

mds: Add Nack message for MDS_TIPC_FCTRL_ENABLED [#3095]




---

** [tickets:#3095] mds: Message loss after splitbrain with enabled 
MDS_TIPC_FCTRL_ENABLED**

**Status:** fixed
**Milestone:** 5.19.10
**Created:** Sat Sep 28, 2019 10:22 PM UTC by Minh Hon Chau
**Last Updated:** Mon Oct 07, 2019 12:14 AM UTC
**Owner:** nobody


After both SCs see each other after split brain, message loss occurs since some 
sequences are missing

2019-09-27T07:13:08.906+02:00 SC-2-1 tipclog[3762]: Lost link 
<1.1.1:eth0-1.1.2:eth0> on network plane A

2019-09-27T07:13:25.369+02:00 SC-2-1 tipclog[3762]: Established link 
<1.1.1:eth0-1.1.2:eth0> on network plane A
2019-09-27T07:13:25.369+02:00 SC-2-1 tipclog[3762]: Established contact with 
node <1.1.2>

mds.log:<139>1 2019-09-27T07:13:25.368051+02:00 SC-2-1 osafamfd 8822 mds.log 
[meta sequenceId="418"] FCTRL: [me] <-- [node:1001002, ref:1599079723], 
RcvData[mseq:8066, mfrag:0, fseq:**539**], rcvwnd[acked:535, rcv:**535**, 
nacked:0], Error[msg loss]

mds.log:<139>1 2019-09-27T07:13:25.378774+02:00 SC-2-1 osafimmd 8684 mds.log 
[meta sequenceId="337"] FCTRL: [me] <-- [node:1001002, ref:2606175488], 
RcvData[mseq:28901, mfrag:0, fseq:**773**], rcvwnd[acked:771, rcv:**771**, 
nacked:0], Error[msg loss]

At this moment, the message loss will be restransmitted in scenario of tipc 
buffer overflow (TIPC_RETDATA), the message loss in this scenario (splitbrain) 
is not retransmitted.

Prior that MDS_TIPC_FCTRL_ENABLED, in this scenario, amfd will forward msg with 
seq=539 to upper layer, and accept to miss the seq=536, 537, 538.
With the fix of this ticket, mds will start the restransmission of 536, 537, 
538.




---

Sent from sourceforge.net because opensaf-tickets@lists.sourceforge.net is 
subscribed to https://sourceforge.net/p/opensaf/tickets/

To unsubscribe from further messages, a project admin can change settings at 
https://sourceforge.net/p/opensaf/admin/tickets/options.  Or, if this is a 
mailing list, you can unsubscribe from the mailing list.___
Opensaf-tickets mailing list
Opensaf-tickets@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets


[tickets] [opensaf:tickets] #3090 mds: Improve tx probation timer for MDS_TIPC_FCTRL_ENABLED

2019-10-14 Thread Minh Hon Chau via Opensaf-tickets
- **status**: assigned --> fixed
- **assigned_to**: Minh Hon Chau -->  nobody 
- **Comment**:

commit 153bb7152691365f787d73ef87e02e34f86eeffc
Author: Minh Chau 
Date:   Fri Oct 11 14:39:28 2019 +1100

mds: Add Intro message [#3090]




---

** [tickets:#3090] mds: Improve tx probation timer for MDS_TIPC_FCTRL_ENABLED**

**Status:** fixed
**Milestone:** 5.19.10
**Created:** Wed Sep 25, 2019 04:18 AM UTC by Minh Hon Chau
**Last Updated:** Wed Sep 25, 2019 04:18 AM UTC
**Owner:** nobody


if MDS_TIPC_FCTRL_ENABLED is enabled only either at sender or receiver (1), the 
tx probation timer will be expired eventually. This tx probation timer 
currently relies on the mds user's data message sent from the peers to 
recognize whethere MDS_TIPC_FCTRL_ENABLED is enabled.
The problem happens, if both peers enable MDS_TIPC_FCTRL_ENABLED, but after the 
sender subscribes the receiver's service and not sending any data until the tx 
probation expires. The sender thinks that the  MDS_TIPC_FCTRL_ENABLED is 
disabled at receiver side and vice versa, hence the mds tipc flow control has 
no effect.
The above problem may not happen within OpenSAF service since most of agent 
will send the Initialize message after mds subscription. However, if the case 
(1) happens, the tx probation prolongs until a data message is sent, which 
should be improved to reduce the tx probation timer.


---

Sent from sourceforge.net because opensaf-tickets@lists.sourceforge.net is 
subscribed to https://sourceforge.net/p/opensaf/tickets/

To unsubscribe from further messages, a project admin can change settings at 
https://sourceforge.net/p/opensaf/admin/tickets/options.  Or, if this is a 
mailing list, you can unsubscribe from the mailing list.___
Opensaf-tickets mailing list
Opensaf-tickets@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets


[tickets] [opensaf:tickets] #3091 mds: Add more overload test cases for mdstest

2019-10-14 Thread Thuan via Opensaf-tickets
- **status**: review --> fixed



---

** [tickets:#3091] mds: Add more overload test cases for mdstest**

**Status:** fixed
**Milestone:** 5.19.10
**Created:** Wed Sep 25, 2019 04:22 AM UTC by Minh Hon Chau
**Last Updated:** Fri Sep 27, 2019 11:03 AM UTC
**Owner:** Thuan


Add more mdstest for ticket #1960. A suggestion:

- mdstest for overload
 . 2 senders overload one receivers
 . one sender overloads 2 receivers

- mdstest for sna
 . without overload, mds sender gradually sends more than 65535 messages, and 
receivers should receive them all
 . with overload, mds sender sends a burst of greater than 65535 messages, and 
receivers should receive them all

- mdstest for #1960 backward compatibility, in order to test the txprob timer
 . sender enables, receiver disables
 . sender enables, receiver enables
 . sender disables, receiver enables

...


---

Sent from sourceforge.net because opensaf-tickets@lists.sourceforge.net is 
subscribed to https://sourceforge.net/p/opensaf/tickets/

To unsubscribe from further messages, a project admin can change settings at 
https://sourceforge.net/p/opensaf/admin/tickets/options.  Or, if this is a 
mailing list, you can unsubscribe from the mailing list.___
Opensaf-tickets mailing list
Opensaf-tickets@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets