From: Stephens, A. <all...@wi...> - 2011-09-13 14:35:41
|
Hi Shridhar: I think the problem you’re seeing is caused by the fact that “tipc-config -be=eth:bond0 -lw=eth:bond0/1000” is being handled by tipc-config as two separate commands – it first does the “be” command and then does the “lw” one. This means that there is a brief period during which TIPC can establish links on “bond0” using the default window size of 50. Later, when the “lw” command finally does execute this tells TIPC to use the larger window size for any new links it sets up over “bond0”, but doesn’t impact the existing links. The easiest solution to your problem would be to change the default window size used for all Ethernet-based links before enabling TIPC on “bond0”; that is, do “tipc-config –lw=eth/1000 –be=eth:bond0”. (Note that this will affect Ethernet links created on other TIPC interfaces, too.) If you only want to use the larger window size on links running over “bond0” then you’ll need to add an additional configuration command that will change the window size for existing links on that interface after you’ve changed the default value used for new links on that interface. This can be done using “tipc-config -be=eth:bond0 -lw=eth:bond0/1000 –lw=?bond0/1000”. Regards, Al From: Shridhar Sahukar [mailto:ssa...@mo...] Sent: Tuesday, September 13, 2011 9:16 AM To: Chandan Kumar A Cc: Stephens, Allan; tip...@li... Subject: RE: [tipc-discussion] Configuring TIPC link window size Hi Chandan, Thanks for the inputs. Yes. I need to have a larger congestion window to handle short lived packet burst. But the problem I am running into is the configuration of window size not getting into effect on all the links. Once it is configured properly I do not see any issues. Btw, I forgot to mention that I am running TIPC version 1.7.7-rc1 Regards, Shridhar On Tue, 2011-09-13 at 09:07 -0400, Chandan Kumar A wrote: Hi Sridhar, I had similar issue during high traffic. So I tried to engineer the link window size (number of unacked packet on send queue). If you increasing the link window size that means the receiving side application is not fast enough to consume the packets or may be your sending side is just too fast. Increasing the link window size could be used, however, it may be able to delay the inevitable if you have continuous high number of packets, and however, it does helps if you are expecting short lived packet bursts. Regards, Chandan -----Original Message----- From: Shridhar Sahukar [mailto:ssa...@mo...] Sent: September-13-11 5:49 AM To: Stephens, Allan Cc: tip...@li...<mailto:tip...@li...> Subject: [tipc-discussion] Configuring TIPC link window size Hi, On our systems TIPC gets configured through following command sequence: modprobe tipc tipc_config -netid=<nid> -a=<address> tipc_config -be=eth:bond0 -lw=eth:bond0/1000 Please note that we want to set the link window size to 1000, so lw option is specified as above. This works most of the time, but intermittently I see that the window size is set to default 50 on some of the links. For example, in the output pasted below from one of machines, you can see that all links have window size 1000 except one of them. Could you please suggest if there is any issue with the way I am configuring the window size? What is the best way to make sure that the window size is configured properly for all the blades? Link <broadcast-link> Window:20 packets RX packets:1875 fragments:0/0 bundles:0/0 TX packets:238 fragments:0/0 bundles:0/0 RX naks:0 defs:0 dups:0 TX naks:0 acks:118 dups:0 Congestion bearer:0 link:0 Send queue max:19 avg:7 Link <1.1.161:bond0-1.1.33:bond0> ACTIVE MTU:1500 Priority:10 Tolerance:1500 ms Window:1000 packets RX packets:29 fragments:0/0 bundles:0/0 TX packets:29 fragments:0/0 bundles:0/0 TX profile sample:2 packets average:64 octets 0-64:100% -256:0% -1024:0% -4096:0% -16354:0% -32768:0% -66000:0% RX states:891 probes:438 naks:0 defs:0 dups:0 TX states:887 probes:438 naks:0 acks:0 dups:0 Congestion bearer:0 link:0 Send queue max:3 avg:0 Link <1.1.161:bond0-1.1.81:bond0> ACTIVE MTU:1500 Priority:10 Tolerance:1500 ms Window:50 packets RX packets:13555 fragments:0/0 bundles:0/0 TX packets:6885 fragments:0/0 bundles:0/0 TX profile sample:220 packets average:134 octets 0-64:2% -256:97% -1024:1% -4096:0% -16354:0% -32768:0% -66000:0% RX states:856 probes:392 naks:0 defs:0 dups:0 TX states:1111 probes:430 naks:0 acks:171 dups:0 Congestion bearer:0 link:0 Send queue max:15 avg:0 Link <1.1.161:bond0-1.1.129:bond0> ACTIVE MTU:1500 Priority:10 Tolerance:1500 ms Window:1000 packets RX packets:29 fragments:0/0 bundles:0/0 TX packets:29 fragments:0/0 bundles:0/0 TX profile sample:2 packets average:64 octets 0-64:100% -256:0% -1024:0% -4096:0% -16354:0% -32768:0% -66000:0% RX states:887 probes:438 naks:0 defs:0 dups:0 TX states:883 probes:434 naks:0 acks:0 dups:0 Congestion bearer:0 link:0 Send queue max:3 avg:0 Link <1.1.161:bond0-1.1.193:bond0> ACTIVE MTU:1500 Priority:10 Tolerance:1500 ms Window:1000 packets RX packets:29 fragments:0/0 bundles:0/0 TX packets:29 fragments:0/0 bundles:0/0 TX profile sample:2 packets average:64 octets 0-64:100% -256:0% -1024:0% -4096:0% -16354:0% -32768:0% -66000:0% RX states:884 probes:435 naks:0 defs:0 dups:0 TX states:880 probes:434 naks:0 acks:0 dups:0 Congestion bearer:0 link:0 Send queue max:3 avg:0 Link <1.1.161:bond0-1.1.225:bond0> ACTIVE MTU:1500 Priority:10 Tolerance:1500 ms Window:1000 packets RX packets:5 fragments:0/0 bundles:0/0 TX packets:5 fragments:0/0 bundles:0/0 TX profile sample:2 packets average:64 octets 0-64:100% -256:0% -1024:0% -4096:0% -16354:0% -32768:0% -66000:0% RX states:888 probes:438 naks:0 defs:0 dups:0 TX states:884 probes:435 naks:0 acks:0 dups:0 Congestion bearer:0 link:0 Send queue max:3 avg:0 Regards, Shridhar ------------------------------------------------------------------------------ BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA Learn about the latest advances in developing for the BlackBerry® mobile platform with sessions, labs & more. See new tools and technologies. Register for BlackBerry® DevCon today! http://p.sf.net/sfu/rim-devcon-copy1 _______________________________________________ tipc-discussion mailing list tip...@li...<mailto:tip...@li...> https://lists.sourceforge.net/lists/listinfo/tipc-discussion =====-----=====-----===== Notice: The information contained in this e-mail message and/or attachments to it may contain confidential or privileged information. If you are not the intended recipient, any dissemination, use, review, distribution, printing or copying of the information contained in this e-mail message and/or attachments to it are strictly prohibited. If you have received this communication in error, please notify us by reply e-mail or telephone and immediately and permanently delete the message and any attachments. Thank you |