docs/colo-proxy.txt

   1 COLO-proxy
   2 ----------
   3 Copyright (c) 2016 Intel Corporation
   4 Copyright (c) 2016 HUAWEI TECHNOLOGIES CO., LTD.
   5 Copyright (c) 2016 Fujitsu, Corp.
   6
   7 This work is licensed under the terms of the GNU GPL, version 2 or later.
   8 See the COPYING file in the top-level directory.
   9
  10 This document gives an overview of COLO proxy's design.
  11
  12 == Background ==
  13 COLO-proxy is a part of COLO project. It is used
  14 to compare the network package to help COLO decide
  15 whether to do checkpoint. With COLO-proxy's help,
  16 COLO greatly improves the performance.
  17
  18 The filter-redirector, filter-mirror, colo-compare
  19 and filter-rewriter compose the COLO-proxy.
  20
  21 == Architecture ==
  22
  23 COLO-Proxy is based on qemu netfilter and it's a plugin for qemu netfilter
  24 (except colo-compare). It keep Secondary VM connect normally to
  25 client and compare packets sent by PVM with sent by SVM.
  26 If the packet difference, notify COLO-frame to do checkpoint and send
  27 all primary packet has queued. Otherwise just send the queued primary
  28 packet and drop the queued secondary packet.
  29
  30 Below is a COLO proxy ascii figure:
  31
  32  Primary qemu                                                           Secondary qemu
  33 +--------------------------------------------------------------+       +----------------------------------------------------------------+
  34 | +----------------------------------------------------------+ |       |  +-----------------------------------------------------------+ |
  35 | |                                                          | |       |  |                                                           | |
  36 | |                        guest                             | |       |  |                        guest                              | |
  37 | |                                                          | |       |  |                                                           | |
  38 | +-------^--------------------------+-----------------------+ |       |  +---------------------+--------+----------------------------+ |
  39 |         |                          |                         |       |                        ^        |                              |
  40 |         |                          |                         |       |                        |        |                              |
  41 |         |  +------------------------------------------------------+  |                        |        |                              |
  42 |netfilter|  |                       |                         |    |  |   netfilter            |        |                              |
  43 | +----------+ +----------------------------+                  |    |  |  +-----------------------------------------------------------+ |
  44 | |       |  |                       |      |        out       |    |  |  |                     |        |  filter excute order       | |
  45 | |       |  |          +-----------------------------+        |    |  |  |                     |        | +------------------->      | |
  46 | |       |  |          |            |      |         |        |    |  |  |                     |        |   TCP                      | |
  47 | | +-----+--+-+  +-----v----+ +-----v----+ |pri +----+----+sec|    |  |  | +------------+  +---+----+---v+rewriter++  +------------+ | |
  48 | | |          |  |          | |          | |in  |         |in |    |  |  | |            |  |        |              |  |            | | |
  49 | | |  filter  |  |  filter  | |  filter  +------>  colo   <------+ +-------->  filter   +--> adjust |   adjust     +-->   filter   | | |
  50 | | |  mirror  |  |redirector| |redirector| |    | compare |   |  |    |  | | redirector |  | ack    |   seq        |  | redirector | | |
  51 | | |          |  |          | |          | |    |         |   |  |    |  | |            |  |        |              |  |            | | |
  52 | | +----^-----+  +----+-----+ +----------+ |    +---------+   |  |    |  | +------------+  +--------+--------------+  +---+--------+ | |
  53 | |      |   tx        |   rx           rx  |                  |  |    |  |            tx                        all       |  rx      | |
  54 | |      |             |                    |                  |  |    |  +-----------------------------------------------------------+ |
  55 | |      |             +--------------+     |                  |  |    |                                                   |            |
  56 | |      |   filter excute order      |     |                  |  |    |                                                   |            |
  57 | |      |  +---------------->        |     |                  |  +--------------------------------------------------------+            |
  58 | +-----------------------------------------+                  |       |                                                                |
  59 |        |                            |                        |       |                                                                |
  60 +--------------------------------------------------------------+       +----------------------------------------------------------------+
  61          |guest receive               | guest send
  62          |                            |
  63 +--------+----------------------------v------------------------+
  64 |                                                              |                          NOTE: filter direction is rx/tx/all
  65 |                         tap                                  |                          rx:receive packets sent to the netdev
  66 |                                                              |                          tx:receive packets sent by the netdev
  67 +--------------------------------------------------------------+
  68
  69 1.Guest receive packet route:
  70
  71 Primary:
  72
  73 Tap --> Mirror Client Filter
  74 Mirror client will send packet to guest,at the
  75 same time, copy and forward packet to secondary
  76 mirror server.
  77
  78 Secondary:
  79
  80 Mirror Server Filter --> TCP Rewriter
  81 If receive packet is TCP packet,we will adjust ack
  82 and update TCP checksum, then send to secondary
  83 guest. Otherwise directly send to guest.
  84
  85 2.Guest send packet route:
  86
  87 Primary:
  88
  89 Guest --> Redirect Server Filter
  90 Redirect server filter receive primary guest packet
  91 but do nothing, just pass to next filter.
  92
  93 Redirect Server Filter --> COLO-Compare
  94 COLO-compare receive primary guest packet then
  95 waiting scondary redirect packet to compare it.
  96 If packet same,send queued primary packet and clear
  97 queued secondary packet, Otherwise send primary packet
  98 and do checkpoint.
  99
 100 COLO-Compare --> Another Redirector Filter
 101 The redirector get packet from colo-compare by use
 102 chardev socket.
 103
 104 Redirector Filter --> Tap
 105 Send the packet.
 106
 107 Secondary:
 108
 109 Guest --> TCP Rewriter Filter
 110 If the packet is TCP packet,we will adjust seq
 111 and update TCP checksum. Then send it to
 112 redirect client filter. Otherwise directly send to
 113 redirect client filter.
 114
 115 Redirect Client Filter --> Redirect Server Filter
 116 Forward packet to primary.
 117
 118 == Components introduction ==
 119
 120 Filter-mirror is a netfilter plugin.
 121 It gives qemu the ability to mirror
 122 packets to a chardev.
 123
 124 Filter-redirector is a netfilter plugin.
 125 It gives qemu the ability to redirect net packet.
 126 Redirector can redirect filter's net packet to outdev,
 127 and redirect indev's packet to filter.
 128
 129                     filter
 130                       +
 131           redirector  |
 132              +--------------+
 133              |        |     |
 134              |        |     |
 135              |        |     |
 136   indev +---------+   +---------->  outdev
 137              |    |         |
 138              |    |         |
 139              |    |         |
 140              +--------------+
 141                   |
 142                   v
 143                 filter
 144
 145 COLO-compare, we do packet comparing job.
 146 Packets coming from the primary char indev will be sent to outdev.
 147 Packets coming from the secondary char dev will be dropped after comparing.
 148 COLO-compare needs two input chardevs and one output chardev:
 149 primary_in=chardev1-id (source: primary send packet)
 150 secondary_in=chardev2-id (source: secondary send packet)
 151 outdev=chardev3-id
 152
 153 Filter-rewriter will rewrite some of secondary packet to make
 154 secondary guest's tcp connection established successfully.
 155 In this module we will rewrite tcp packet's ack to the secondary
 156 from primary,and rewrite tcp packet's seq to the primary from
 157 secondary.
 158
 159 == Usage ==
 160
 161 Here is an example using demonstration IP and port addresses to more
 162 clearly describe the usage.
 163
 164 Primary(ip:3.3.3.3):
 165 -netdev tap,id=hn0,vhost=off,script=/etc/qemu-ifup,downscript=/etc/qemu-ifdown
 166 -device e1000,id=e0,netdev=hn0,mac=52:a4:00:12:78:66
 167 -chardev socket,id=mirror0,host=3.3.3.3,port=9003,server,nowait
 168 -chardev socket,id=compare1,host=3.3.3.3,port=9004,server,nowait
 169 -chardev socket,id=compare0,host=3.3.3.3,port=9001,server,nowait
 170 -chardev socket,id=compare0-0,host=3.3.3.3,port=9001
 171 -chardev socket,id=compare_out,host=3.3.3.3,port=9005,server,nowait
 172 -chardev socket,id=compare_out0,host=3.3.3.3,port=9005
 173 -object iothread,id=iothread1
 174 -object filter-mirror,id=m0,netdev=hn0,queue=tx,outdev=mirror0
 175 -object filter-redirector,netdev=hn0,id=redire0,queue=rx,indev=compare_out
 176 -object filter-redirector,netdev=hn0,id=redire1,queue=rx,outdev=compare0
 177 -object colo-compare,id=comp0,primary_in=compare0-0,secondary_in=compare1,outdev=compare_out0,iothread=iothread1
 178
 179 Secondary(ip:3.3.3.8):
 180 -netdev tap,id=hn0,vhost=off,script=/etc/qemu-ifup,down script=/etc/qemu-ifdown
 181 -device e1000,netdev=hn0,mac=52:a4:00:12:78:66
 182 -chardev socket,id=red0,host=3.3.3.3,port=9003
 183 -chardev socket,id=red1,host=3.3.3.3,port=9004
 184 -object filter-redirector,id=f1,netdev=hn0,queue=tx,indev=red0
 185 -object filter-redirector,id=f2,netdev=hn0,queue=rx,outdev=red1
 186 -object filter-rewriter,id=f3,netdev=hn0,queue=all
 187
 188 If you want to use virtio-net-pci or other driver with vnet_header:
 189
 190 Primary(ip:3.3.3.3):
 191 -netdev tap,id=hn0,vhost=off,script=/etc/qemu-ifup,downscript=/etc/qemu-ifdown
 192 -device e1000,id=e0,netdev=hn0,mac=52:a4:00:12:78:66
 193 -chardev socket,id=mirror0,host=3.3.3.3,port=9003,server,nowait
 194 -chardev socket,id=compare1,host=3.3.3.3,port=9004,server,nowait
 195 -chardev socket,id=compare0,host=3.3.3.3,port=9001,server,nowait
 196 -chardev socket,id=compare0-0,host=3.3.3.3,port=9001
 197 -chardev socket,id=compare_out,host=3.3.3.3,port=9005,server,nowait
 198 -chardev socket,id=compare_out0,host=3.3.3.3,port=9005
 199 -object filter-mirror,id=m0,netdev=hn0,queue=tx,outdev=mirror0,vnet_hdr_support
 200 -object filter-redirector,netdev=hn0,id=redire0,queue=rx,indev=compare_out,vnet_hdr_support
 201 -object filter-redirector,netdev=hn0,id=redire1,queue=rx,outdev=compare0,vnet_hdr_support
 202 -object colo-compare,id=comp0,primary_in=compare0-0,secondary_in=compare1,outdev=compare_out0,vnet_hdr_support
 203
 204 Secondary(ip:3.3.3.8):
 205 -netdev tap,id=hn0,vhost=off,script=/etc/qemu-ifup,down script=/etc/qemu-ifdown
 206 -device e1000,netdev=hn0,mac=52:a4:00:12:78:66
 207 -chardev socket,id=red0,host=3.3.3.3,port=9003
 208 -chardev socket,id=red1,host=3.3.3.3,port=9004
 209 -object filter-redirector,id=f1,netdev=hn0,queue=tx,indev=red0,vnet_hdr_support
 210 -object filter-redirector,id=f2,netdev=hn0,queue=rx,outdev=red1,vnet_hdr_support
 211 -object filter-rewriter,id=f3,netdev=hn0,queue=all,vnet_hdr_support
 212
 213 Note:
 214   a.COLO-proxy must work with COLO-frame and Block-replication.
 215   b.Primary COLO must be started firstly, because COLO-proxy needs
 216     chardev socket server running before secondary started.
 217   c.Filter-rewriter only rewrite tcp packet.