> (...)
>
> I think I've found the reason now and require confirmation from the maillist
> now. It looks like that polyclt will refresh the ifconfig (inet 10.101.128.1
> netmask...etc.) the polysrv has written, doesn't it?
-------
I see. Sorry, it's only a tedious problem. There is an option
'delete_old_addrs 0' can be used to prevent old alias being deleted.
However, new problem has occured in my single-machine PolyMix4 attempt.
(--proxy is still blank yet)
=================================
The modifications I've done on "my-polymix-4.pg" are as below:
a) client_side and server_side:
hosts = [ '192.168.241.6' ]; // replace with IPs of your PCs
b) TheBench.peak_req_rate = 75/sec;
c) rate FillRate = 75%*TheBench.peak_req_rate;
d) size ProxyCacheSize = 9GB;
e) Resolver.servers = ['192.168.241.6:53' ]; // named is running.
Although I'm not sure if PolyMix4 can be run on single machine
(polysrv&polyclt), I've done step by step according to PolyMix4's how-to.
(Except that, netperf is not required; and ipfw is not used)
FreeBSD's kernel has not been recompiled, but --
MSL is set to 3s; portrange is set to 40000;
Struggling with the PolyMix4 til now~, yet I still met with the following
error:
----------------------------------------------
000.05| created PolyMix-4-rbt [187 / 06f37487.305f0d09:00000586] on 10.101.0.94
000.05| created PolyMix-4-rbt [188 / 06f37487.305f0d09:00000588] on 10.101.0.94
000.05| created 188 agents total
000.05| fyi: current state (1) stored
000.05| fyi: working set size goal: first 388800 fill objects of the test
000.05| fyi: max local population size: 188 robots
000.13| i-framp 10 2.00 1781 0.00 0 19
000.22| i-framp 34 4.79 2439 0.00 0 19
000.30| i-framp 47 2.59 2983 0.00 0 20
000.38| i-framp 57 2.00 2009 0.00 0 26
000.47| i-framp 73 3.19 2416 0.00 0 25
000.55| i-framp 85 2.39 2506 0.00 0 21
000.63| i-framp 96 2.20 2752 0.00 0 29
000.72| i-framp 113 3.39 2858 0.00 0 26
000.80| i-framp 121 1.60 2570 0.00 0 28
000.81| fyi: servers visible: 519 requests sent: 58 servers ready-to-be-hit: 56 ( 10.79%)
000.88| i-framp 136 2.99 2389 0.00 0 37
000.97| i-framp 173 7.38 2284 0.00 0 38
000.97| Connection.cc:96: error: 1/1 (s55) No buffer space available
000.97| Client.cc:265: error: 1/2 (c53) failed to establish a connection
000.97| 10.101.0.87 failed to connect to 10.101.129.209:8000
000.97| Connection.cc:96: error: 2/3 (s55) No buffer space available
000.97| Client.cc:265: error: 2/4 (c53) failed to establish a connection
000.97| 10.101.0.87 failed to connect to 10.101.129.209:8000
000.99| ExpPortMgr.cc:150: error: 1/7 (s48) Address already in use
000.99| failed to bind to 10.101.0.27:9117
001.01| Connection.cc:96: error: 4/8 (s55) No buffer space available
001.01| Client.cc:265: error: 4/9 (c53) failed to establish a connection
001.01| 10.101.0.56 failed to connect to 10.101.129.18:8000
001.01| Connection.cc:96: error: 8/16 (s55) No buffer space available
001.01| Client.cc:265: error: 8/17 (c53) failed to establish a connection
001.01| 10.101.0.56 failed to connect to 10.101.129.18:8000
001.04| Connection.cc:173: error: 1/28 (s54) Connection reset by peer
001.04| error reading from 10.101.128.234:8000 after 1 reads, 1 writes, 1 xacts
001.05| i-framp 203 6.19 2630 0.00 14 33
001.08| Connection.cc:173: error: 2/29 (s54) Connection reset by peer
001.08| error reading from 10.101.128.97:8000 after 1 reads, 1 writes, 1 xacts
001.13| Connection.cc:173: error: 4/35 (s54) Connection reset by peer
001.13| error reading from 10.101.128.86:8000 after 1 reads, 1 writes, 1 xacts
001.14| i-framp 222 4.39 2800 0.00 5 35
001.22| i-framp 243 4.19 2520 0.00 0 41
001.22| Connection.cc:96: error: 16/36 (s55) No buffer space available
001.22| Client.cc:265: error: 16/37 (c53) failed to establish a connection
001.22| 10.101.0.87 failed to connect to 10.101.129.26:8000
001.30| i-framp 263 3.99 2063 0.00 1 45
----------------------------------------------
Is it mean I need to increase max mbuf clusters to 32k by tuning the kernel?
Or, it's in fact impossible to run polyclt&srv with PolyMix-4.pg in single
machine? Test is to be continued.
tchu
> In fact, I need to compare different Cache-Peering(Hierarchy or Cache Digest)
> efficiency in my project. At least 3 cache proxies is needed, if the newest
> PolyMix4 cannot be used, which workload will be recommended then?
> thanks in advance'
>
> Tiancheng Hu
> Kanazawa Institute of Technology, Japan.
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:24 MST