| blob | ad24680d130e86734d962c6a2701392862d88ba7 |
1 COarse-grained LOck-stepping Virtual Machines for Non-stop Service
2 ----------------------------------------
3 Copyright (c) 2016 Intel Corporation
4 Copyright (c) 2016 HUAWEI TECHNOLOGIES CO., LTD.
5 Copyright (c) 2016 Fujitsu, Corp.
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.
10 This document gives an overview of COLO's design and how to use it.
12 == Background ==
13 Virtual machine (VM) replication is a well known technique for providing
14 application-agnostic software-implemented hardware fault tolerance,
15 also known as "non-stop service".
17 COLO (COarse-grained LOck-stepping) is a high availability solution.
18 Both primary VM (PVM) and secondary VM (SVM) run in parallel. They receive the
19 same request from client, and generate response in parallel too.
20 If the response packets from PVM and SVM are identical, they are released
21 immediately. Otherwise, a VM checkpoint (on demand) is conducted.
23 == Architecture ==
25 The architecture of COLO is shown in the diagram below.
26 It consists of a pair of networked physical nodes:
27 The primary node running the PVM, and the secondary node running the SVM
28 to maintain a valid replica of the PVM.
29 PVM and SVM execute in parallel and generate output of response packets for
30 client requests according to the application semantics.
32 The incoming packets from the client or external network are received by the
33 primary node, and then forwarded to the secondary node, so that both the PVM
34 and the SVM are stimulated with the same requests.
36 COLO receives the outbound packets from both the PVM and SVM and compares them
37 before allowing the output to be sent to clients.
39 The SVM is qualified as a valid replica of the PVM, as long as it generates
40 identical responses to all client requests. Once the differences in the outputs
41 are detected between the PVM and SVM, COLO withholds transmission of the
42 outbound packets until it has successfully synchronized the PVM state to the SVM.
44 Primary Node Secondary Node
45 +------------+ +-----------------------+ +------------------------+ +------------+
46 | | | HeartBeat +<----->+ HeartBeat | | |
47 | Primary VM | +-----------+-----------+ +-----------+------------+ |Secondary VM|
48 | | | | | |
49 | | +-----------|-----------+ +-----------|------------+ | |
50 | | |QEMU +---v----+ | |QEMU +----v---+ | | |
51 | | | |Failover| | | |Failover| | | |
52 | | | +--------+ | | +--------+ | | |
53 | | | +---------------+ | | +---------------+ | | |
54 | | | | VM Checkpoint +-------------->+ VM Checkpoint | | | |
55 | | | +---------------+ | | +---------------+ | | |
56 |Requests<--------------------------\ /-----------------\ /--------------------->Requests|
57 | | | ^ ^ | | | | | | |
58 |Responses+---------------------\ /-|-|------------\ /-------------------------+Responses|
59 | | | | | | | | | | | | | | | |
60 | | | +-----------+ | | | | | | | | | | +----------+ | | |
61 | | | | COLO disk | | | | | | | | | | | | COLO disk| | | |
62 | | | | Manager +---------------------------->| Manager | | | |
63 | | | ++----------+ v v | | | | | v v | +---------++ | | |
64 | | | |+-----------+-+-+-++| | ++-+--+-+---------+ | | | |
65 | | | || COLO Proxy || | | COLO Proxy | | | | |
66 | | | || (compare packet || | |(adjust sequence | | | | |
67 | | | ||and mirror packet)|| | | and ACK) | | | | |
68 | | | |+------------+---+-+| | +-----------------+ | | | |
69 +------------+ +-----------------------+ +------------------------+ +------------+
70 +------------+ | | | | +------------+
71 | VM Monitor | | | | | | VM Monitor |
72 +------------+ | | | | +------------+
73 +---------------------------------------+ +----------------------------------------+
74 | Kernel | | | | | Kernel | |
75 +---------------------------------------+ +----------------------------------------+
76 | | | |
77 +--------------v+ +---------v---+--+ +------------------+ +v-------------+
78 | Storage | |External Network| | External Network | | Storage |
79 +---------------+ +----------------+ +------------------+ +--------------+
82 == Components introduction ==
84 You can see there are several components in COLO's diagram of architecture.
85 Their functions are described below.
87 HeartBeat:
88 Runs on both the primary and secondary nodes, to periodically check platform
89 availability. When the primary node suffers a hardware fail-stop failure,
90 the heartbeat stops responding, the secondary node will trigger a failover
91 as soon as it determines the absence.
93 COLO disk Manager:
94 When primary VM writes data into image, the colo disk manger captures this data
95 and sends it to secondary VM's which makes sure the context of secondary VM's
96 image is consistent with the context of primary VM 's image.
97 For more details, please refer to docs/block-replication.txt.
99 Checkpoint/Failover Controller:
100 Modifications of save/restore flow to realize continuous migration,
101 to make sure the state of VM in Secondary side is always consistent with VM in
102 Primary side.
104 COLO Proxy:
105 Delivers packets to Primary and Secondary, and then compare the responses from
106 both side. Then decide whether to start a checkpoint according to some rules.
107 Please refer to docs/colo-proxy.txt for more information.
109 Note:
110 HeartBeat has not been implemented yet, so you need to trigger failover process
111 by using 'x-colo-lost-heartbeat' command.
113 == COLO operation status ==
115 +-----------------+
116 | |
117 | Start COLO |
118 | |
119 +--------+--------+
120 |
121 | Main qmp command:
122 | migrate-set-capabilities with x-colo
123 | migrate
124 |
125 v
126 +--------+--------+
127 | |
128 | COLO running |
129 | |
130 +--------+--------+
131 |
132 | Main qmp command:
133 | x-colo-lost-heartbeat
134 | or
135 | some error happened
136 v
137 +--------+--------+
138 | | send qmp event:
139 | COLO failover | COLO_EXIT
140 | |
141 +-----------------+
143 COLO use the qmp command to switch and report operation status.
144 The diagram just shows the main qmp command, you can get the detail
145 in test procedure.
147 == Test procedure ==
148 1. Startup qemu
149 Primary:
150 # qemu-system-x86_64 -accel kvm -m 2048 -smp 2 -qmp stdio -name primary \
151 -device piix3-usb-uhci -vnc :7 \
152 -device usb-tablet -netdev tap,id=hn0,vhost=off \
153 -device virtio-net-pci,id=net-pci0,netdev=hn0 \
154 -drive if=virtio,id=primary-disk0,driver=quorum,read-pattern=fifo,vote-threshold=1,\
155 children.0.file.filename=1.raw,\
156 children.0.driver=raw -S
157 Secondary:
158 # qemu-system-x86_64 -accel kvm -m 2048 -smp 2 -qmp stdio -name secondary \
159 -device piix3-usb-uhci -vnc :7 \
160 -device usb-tablet -netdev tap,id=hn0,vhost=off \
161 -device virtio-net-pci,id=net-pci0,netdev=hn0 \
162 -drive if=none,id=secondary-disk0,file.filename=1.raw,driver=raw,node-name=node0 \
163 -drive if=virtio,id=active-disk0,driver=replication,mode=secondary,\
164 file.driver=qcow2,top-id=active-disk0,\
165 file.file.filename=/mnt/ramfs/active_disk.img,\
166 file.backing.driver=qcow2,\
167 file.backing.file.filename=/mnt/ramfs/hidden_disk.img,\
168 file.backing.backing=secondary-disk0 \
169 -incoming tcp:0:8888
171 2. On Secondary VM's QEMU monitor, issue command
172 {'execute':'qmp_capabilities'}
173 { 'execute': 'nbd-server-start',
174 'arguments': {'addr': {'type': 'inet', 'data': {'host': 'xx.xx.xx.xx', 'port': '8889'} } }
175 }
176 {'execute': 'nbd-server-add', 'arguments': {'device': 'secondary-disk0', 'writable': true } }
178 Note:
179 a. The qmp command nbd-server-start and nbd-server-add must be run
180 before running the qmp command migrate on primary QEMU
181 b. Active disk, hidden disk and nbd target's length should be the
182 same.
183 c. It is better to put active disk and hidden disk in ramdisk.
185 3. On Primary VM's QEMU monitor, issue command:
186 {'execute':'qmp_capabilities'}
187 { 'execute': 'human-monitor-command',
188 'arguments': {'command-line': 'drive_add -n buddy driver=replication,mode=primary,file.driver=nbd,file.host=xx.xx.xx.xx,file.port=8889,file.export=secondary-disk0,node-name=nbd_client0'}}
189 { 'execute':'x-blockdev-change', 'arguments':{'parent': 'primary-disk0', 'node': 'nbd_client0' } }
190 { 'execute': 'migrate-set-capabilities',
191 'arguments': {'capabilities': [ {'capability': 'x-colo', 'state': true } ] } }
192 { 'execute': 'migrate', 'arguments': {'uri': 'tcp:xx.xx.xx.xx:8888' } }
194 Note:
195 a. There should be only one NBD Client for each primary disk.
196 b. xx.xx.xx.xx is the secondary physical machine's hostname or IP
197 c. The qmp command line must be run after running qmp command line in
198 secondary qemu.
200 4. After the above steps, you will see, whenever you make changes to PVM, SVM will be synced.
201 You can issue command '{ "execute": "migrate-set-parameters" , "arguments":{ "x-checkpoint-delay": 2000 } }'
202 to change the checkpoint period time
204 5. Failover test
205 You can kill Primary VM and run 'x_colo_lost_heartbeat' in Secondary VM's
206 monitor at the same time, then SVM will failover and client will not detect this
207 change.
209 Before issuing '{ "execute": "x-colo-lost-heartbeat" }' command, we have to
210 issue block related command to stop block replication.
211 Primary:
212 Remove the nbd child from the quorum:
213 { 'execute': 'x-blockdev-change', 'arguments': {'parent': 'colo-disk0', 'child': 'children.1'}}
214 { 'execute': 'human-monitor-command','arguments': {'command-line': 'drive_del blk-buddy0'}}
215 Note: there is no qmp command to remove the blockdev now
217 Secondary:
218 The primary host is down, so we should do the following thing:
219 { 'execute': 'nbd-server-stop' }
221 == TODO ==
222 1. Support continuous VM replication.
223 2. Support shared storage.
224 3. Develop the heartbeat part.
225 4. Reduce checkpoint VM’s downtime while doing checkpoint.