[Spread-users] high cpu load
Theodore D. Hickman Jr.
ted.hickman at prohosters.com
Mon Aug 27 10:53:07 EDT 2001
And me, because I have the EXACT same problem with CPU usage.
And what's even more frustrating is that spread dies every time I get
LARGE amounts of traffic, like in excess of 40Mb/s across the cluster.
Theodore D. Hickman Jr.
Pro Hosters L.L.C.
Monitor>
============================
Status at scandium.4ph.com V 3.15. 2 (state 1, gstate 1) after 4425
seconds :
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 875937 tok_hurry : 6 memb change: 2
sent pack: 319661 recv pack : 1110446 retrans : 7590
u retrans: 7519 s retrans : 71 b retrans : 0
My_aru : 2049140 Aru : 2049134 Highest seq: 2049140
Sessions : 225 Groups : 20 Window : 60
Deliver M: 2446092 Deliver Pk: 2471654 Pers Window: 15
Delta Mes: -640779 Delta Pack: -1 Delta sec : -1253
==================================
Monitor>
============================
Status at zirconium.4ph.com V 3.15. 2 (state 1, gstate 1) after 18369
seconds :
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 6042732 tok_hurry : 83 memb change: 18
sent pack: 938850 recv pack : 4887768 retrans : 2396
u retrans: 2129 s retrans : 267 b retrans : 0
My_aru : 2049140 Aru : 2049136 Highest seq: 2049140
Sessions : 194 Groups : 66 Window : 60
Deliver M: 8403786 Deliver Pk: 8505193 Pers Window: 15
Delta Mes: 5957694 Delta Pack: 2 Delta sec : 13944
==================================
Monitor>
============================
Status at neon.4ph.com V 3.15. 2 (state 1, gstate 1) after 5625 seconds
:
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 1156997 tok_hurry : 6 memb change: 6
sent pack: 295430 recv pack : 1531728 retrans : 540
u retrans: 540 s retrans : 0 b retrans : 0
My_aru : 2049140 Aru : 2049133 Highest seq: 2049140
Sessions : 202 Groups : 20 Window : 60
Deliver M: 3075890 Deliver Pk: 3108191 Pers Window: 15
Delta Mes: -5327896 Delta Pack: -3 Delta sec : -12744
==================================
Monitor>
============================
Status at titanium.4ph.com V 3.15. 2 (state 1, gstate 1) after 18357
seconds :
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 6035237 tok_hurry : 24 memb change: 16
sent pack: 1308609 recv pack : 4508983 retrans : 2347
u retrans: 2144 s retrans : 203 b retrans : 0
My_aru : 2066550 Aru : 2066545 Highest seq: 2066550
Sessions : 216 Groups : 20 Window : 60
Deliver M: 8406602 Deliver Pk: 8507902 Pers Window: 15
Delta Mes: 5302520 Delta Pack: 2 Delta sec : 12649
==================================
Monitor>
============================
Status at platinum.4ph.com V 3.15. 2 (state 1, gstate 1) after 5718
seconds :
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 1175632 tok_hurry : 7 memb change: 7
sent pack: 416735 recv pack : 1423012 retrans : 16195
u retrans: 14144 s retrans : 2051 b retrans : 0
My_aru : 2072452 Aru : 2072440 Highest seq: 2072452
Sessions : 226 Groups : 20 Window : 60
Deliver M: 3109930 Deliver Pk: 3142618 Pers Window: 15
Delta Mes: 10981 Delta Pack: -5 Delta sec : 53
==================================
Monitor> Monitor: send status query
============================
Status at gold.4ph.com V 3.15. 2 (state 1, gstate 1) after 18459 seconds
:
Membership : 6 procs in 1 segments, leader is zirconium.4ph.com
rounds : 6054195 tok_hurry : 23 memb change: 18
sent pack: 341 recv pack : 5839322 retrans : 17258
u retrans: 17219 s retrans : 39 b retrans : 0
My_aru : 2095431 Aru : 2095428 Highest seq: 2095431
Sessions : 1 Groups : 66 Window : 60
Deliver M: 8449573 Deliver Pk: 8551484 Pers Window: 15
Delta Mes: 5162 Delta Pack: 5221 Delta sec : 20
==================================
-----Original Message-----
From: spread-users-admin at lists.spread.org
[mailto:spread-users-admin at lists.spread.org] On Behalf Of Yair Amir
Sent: Monday, August 27, 2001 10:54 AM
To: Dirk Vleugels
Cc: spread-users at lists.spread.org; Michael Stiller
Subject: Re: [Spread-users] high cpu load
Hi,
Something seems wrong with your network. I think Spread should use less
than 10% cpu.
You have way to many retransmissions. You have only about 30 packets/sec
which is really
low. Usually in speeds below thousands of packets/sec, there should
almost be no retransmissions.
I don't know why this happens on your network.
Cheers,
:) Yair.
Dirk Vleugels wrote:
>
> Hi,
>
> we're using mod_log_spread to combine the access logs of a 4-node LVS
> cluster. The CPU usage of the spread daemons (on the apache
> real-servers) seems quite high (up to 60%), while the loghost is
> idle.
>
> spmonitor shows the following numbers:
>
> Status at cluster1 V 3.16. 0 (state 1, gstate 1) after 760918 seconds
:
> Membership : 5 procs in 1 segments, leader is cluster1
> rounds : 617634183 tok_hurry : 2144198 memb change: 5
> sent pack: 5706975 recv pack : 17722855 retrans : 5540933
> u retrans: 5386778 s retrans : 154155 b retrans : 0
> My_aru : 14529311 Aru : 14529311 Highest seq: 14529311
> Sessions : 163 Groups : 3 Window : 60
> Deliver M: 24034352 Deliver Pk: 24133000 Pers Window: 15
> Delta Mes: 224840 Delta Pack: 0 Delta sec : 129196
> ==================================
>
> Monitor>
> ============================
> Status at cluster3 V 3.16. 0 (state 1, gstate 1) after 761407 seconds
:
> Membership : 5 procs in 1 segments, leader is cluster1
> rounds : 617636072 tok_hurry : 2087959 memb change: 9
> sent pack: 5637323 recv pack : 23049662 retrans : 5616992
> u retrans: 5557058 s retrans : 59934 b retrans : 0
> My_aru : 14529311 Aru : 14529311 Highest seq: 14529311
> Sessions : 128 Groups : 3 Window : 60
> Deliver M: 24034364 Deliver Pk: 24133022 Pers Window: 15
> Delta Mes: 12 Delta Pack: 0 Delta sec : 489
> ==================================
>
> Monitor>
> ============================
> Status at cluster2 V 3.16. 0 (state 1, gstate 1) after 761172 seconds
:
> Membership : 5 procs in 1 segments, leader is cluster1
> rounds : 617635235 tok_hurry : 2087845 memb change: 7
> sent pack: 5329799 recv pack : 23207452 retrans : 603931
> u retrans: 416 s retrans : 603515 b retrans : 0
> My_aru : 14529311 Aru : 14529311 Highest seq: 14529311
> Sessions : 148 Groups : 3 Window : 60
> Deliver M: 24034359 Deliver Pk: 24133012 Pers Window: 15
> Delta Mes: -5 Delta Pack: 0 Delta sec : -235
> ==================================
>
> Monitor>
> ============================
> Status at cluster4 V 3.16. 0 (state 1, gstate 1) after 761531 seconds
:
> Membership : 5 procs in 1 segments, leader is cluster1
> rounds : 617637065 tok_hurry : 2088017 memb change: 11
> sent pack: 5715536 recv pack : 22496807 retrans : 5397307
> u retrans: 5163263 s retrans : 234044 b retrans : 0
> My_aru : 14529311 Aru : 14529311 Highest seq: 14529311
> Sessions : 148 Groups : 3 Window : 60
> Deliver M: 24034370 Deliver Pk: 24133048 Pers Window: 15
> Delta Mes: 11 Delta Pack: 0 Delta sec : 359
>
> Monitor>
> ============================
> Status at loghost V 3.16. 0 (state 1, gstate 1) after 631721 seconds :
> Membership : 5 procs in 1 segments, leader is cluster1
> rounds : 595498478 tok_hurry : 1867064 memb change: 3
> sent pack: 2 recv pack : 28541855 retrans : 4951882
> u retrans: 4809775 s retrans : 142107 b retrans : 0
> My_aru : 14529246 Aru : 14529246 Highest seq: 14529246
> Sessions : 1 Groups : 3 Window : 60
> Deliver M: 23809447 Deliver Pk: 23906883 Pers Window: 15
> Delta Mes: -224800 Delta Pack: 58 Delta sec : -129808
> ==================================
>
> Any hints to reduce the cpu load? Something misconfigured? What
> puzzles me are the membership changes. Afaik, only the first apache
> instance does the logging, so the start of more httpd's shouldn't
> affect this counter.
>
> Regards,
> Dirk
>
> --
> Dirk Vleugels E-Mail: dvl at 2scale.net
> 2scale GmbH Phone: +49 231 28 661 52
> Am Rosenplätzchen 20 Fax: +49 231 28 661 59
> 44269 Dortmund, Germany Mobile: +49 173 28 106 61
>
> _______________________________________________
> spread-users mailing list
> spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
_______________________________________________
spread-users mailing list
spread-users at lists.spread.org
http://lists.spread.org/mailman/listinfo/spread-users
More information about the Spread-users
mailing list