]>
Commit | Line | Data |
---|---|---|
1 | ifdef::manvolnum[] | |
2 | PVE(8) | |
3 | ====== | |
4 | include::attributes.txt[] | |
5 | ||
6 | :pve-toplevel: | |
7 | ||
8 | NAME | |
9 | ---- | |
10 | ||
11 | pmxcfs - Proxmox Cluster File System | |
12 | ||
13 | SYNOPSIS | |
14 | -------- | |
15 | ||
16 | include::pmxcfs.8-cli.adoc[] | |
17 | ||
18 | DESCRIPTION | |
19 | ----------- | |
20 | endif::manvolnum[] | |
21 | ||
22 | ifndef::manvolnum[] | |
23 | Proxmox Cluster File System (pmxcfs) | |
24 | ==================================== | |
25 | include::attributes.txt[] | |
26 | endif::manvolnum[] | |
27 | ||
28 | ifdef::wiki[] | |
29 | :pve-toplevel: | |
30 | endif::wiki[] | |
31 | ||
32 | The Proxmox Cluster file system (``pmxcfs'') is a database-driven file | |
33 | system for storing configuration files, replicated in real time to all | |
34 | cluster nodes using `corosync`. We use this to store all PVE related | |
35 | configuration files. | |
36 | ||
37 | Although the file system stores all data inside a persistent database | |
38 | on disk, a copy of the data resides in RAM. That imposes restriction | |
39 | on the maximum size, which is currently 30MB. This is still enough to | |
40 | store the configuration of several thousand virtual machines. | |
41 | ||
42 | This system provides the following advantages: | |
43 | ||
44 | * seamless replication of all configuration to all nodes in real time | |
45 | * provides strong consistency checks to avoid duplicate VM IDs | |
46 | * read-only when a node loses quorum | |
47 | * automatic updates of the corosync cluster configuration to all nodes | |
48 | * includes a distributed locking mechanism | |
49 | ||
50 | ||
51 | POSIX Compatibility | |
52 | ------------------- | |
53 | ||
54 | The file system is based on FUSE, so the behavior is POSIX like. But | |
55 | some feature are simply not implemented, because we do not need them: | |
56 | ||
57 | * you can just generate normal files and directories, but no symbolic | |
58 | links, ... | |
59 | ||
60 | * you can't rename non-empty directories (because this makes it easier | |
61 | to guarantee that VMIDs are unique). | |
62 | ||
63 | * you can't change file permissions (permissions are based on path) | |
64 | ||
65 | * `O_EXCL` creates were not atomic (like old NFS) | |
66 | ||
67 | * `O_TRUNC` creates are not atomic (FUSE restriction) | |
68 | ||
69 | ||
70 | File Access Rights | |
71 | ------------------ | |
72 | ||
73 | All files and directories are owned by user `root` and have group | |
74 | `www-data`. Only root has write permissions, but group `www-data` can | |
75 | read most files. Files below the following paths: | |
76 | ||
77 | /etc/pve/priv/ | |
78 | /etc/pve/nodes/${NAME}/priv/ | |
79 | ||
80 | are only accessible by root. | |
81 | ||
82 | ||
83 | Technology | |
84 | ---------- | |
85 | ||
86 | We use the http://www.corosync.org[Corosync Cluster Engine] for | |
87 | cluster communication, and http://www.sqlite.org[SQlite] for the | |
88 | database file. The file system is implemented in user space using | |
89 | http://fuse.sourceforge.net[FUSE]. | |
90 | ||
91 | File System Layout | |
92 | ------------------ | |
93 | ||
94 | The file system is mounted at: | |
95 | ||
96 | /etc/pve | |
97 | ||
98 | Files | |
99 | ~~~~~ | |
100 | ||
101 | [width="100%",cols="m,d"] | |
102 | |======= | |
103 | |`corosync.conf` | Corosync cluster configuration file (previous to {pve} 4.x this file was called cluster.conf) | |
104 | |`storage.cfg` | {pve} storage configuration | |
105 | |`datacenter.cfg` | {pve} datacenter wide configuration (keyboard layout, proxy, ...) | |
106 | |`user.cfg` | {pve} access control configuration (users/groups/...) | |
107 | |`domains.cfg` | {pve} authentication domains | |
108 | |`authkey.pub` | Public key used by ticket system | |
109 | |`pve-root-ca.pem` | Public certificate of cluster CA | |
110 | |`priv/shadow.cfg` | Shadow password file | |
111 | |`priv/authkey.key` | Private key used by ticket system | |
112 | |`priv/pve-root-ca.key` | Private key of cluster CA | |
113 | |`nodes/<NAME>/pve-ssl.pem` | Public SSL certificate for web server (signed by cluster CA) | |
114 | |`nodes/<NAME>/pve-ssl.key` | Private SSL key for `pve-ssl.pem` | |
115 | |`nodes/<NAME>/pveproxy-ssl.pem` | Public SSL certificate (chain) for web server (optional override for `pve-ssl.pem`) | |
116 | |`nodes/<NAME>/pveproxy-ssl.key` | Private SSL key for `pveproxy-ssl.pem` (optional) | |
117 | |`nodes/<NAME>/qemu-server/<VMID>.conf` | VM configuration data for KVM VMs | |
118 | |`nodes/<NAME>/lxc/<VMID>.conf` | VM configuration data for LXC containers | |
119 | |`firewall/cluster.fw` | Firewall configuration applied to all nodes | |
120 | |`firewall/<NAME>.fw` | Firewall configuration for individual nodes | |
121 | |`firewall/<VMID>.fw` | Firewall configuration for VMs and Containers | |
122 | |======= | |
123 | ||
124 | ||
125 | Symbolic links | |
126 | ~~~~~~~~~~~~~~ | |
127 | ||
128 | [width="100%",cols="m,m"] | |
129 | |======= | |
130 | |`local` | `nodes/<LOCAL_HOST_NAME>` | |
131 | |`qemu-server` | `nodes/<LOCAL_HOST_NAME>/qemu-server/` | |
132 | |`lxc` | `nodes/<LOCAL_HOST_NAME>/lxc/` | |
133 | |======= | |
134 | ||
135 | ||
136 | Special status files for debugging (JSON) | |
137 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
138 | ||
139 | [width="100%",cols="m,d"] | |
140 | |======= | |
141 | |`.version` |File versions (to detect file modifications) | |
142 | |`.members` |Info about cluster members | |
143 | |`.vmlist` |List of all VMs | |
144 | |`.clusterlog` |Cluster log (last 50 entries) | |
145 | |`.rrd` |RRD data (most recent entries) | |
146 | |======= | |
147 | ||
148 | ||
149 | Enable/Disable debugging | |
150 | ~~~~~~~~~~~~~~~~~~~~~~~~ | |
151 | ||
152 | You can enable verbose syslog messages with: | |
153 | ||
154 | echo "1" >/etc/pve/.debug | |
155 | ||
156 | And disable verbose syslog messages with: | |
157 | ||
158 | echo "0" >/etc/pve/.debug | |
159 | ||
160 | ||
161 | Recovery | |
162 | -------- | |
163 | ||
164 | If you have major problems with your Proxmox VE host, e.g. hardware | |
165 | issues, it could be helpful to just copy the pmxcfs database file | |
166 | `/var/lib/pve-cluster/config.db` and move it to a new Proxmox VE | |
167 | host. On the new host (with nothing running), you need to stop the | |
168 | `pve-cluster` service and replace the `config.db` file (needed permissions | |
169 | `0600`). Second, adapt `/etc/hostname` and `/etc/hosts` according to the | |
170 | lost Proxmox VE host, then reboot and check. (And don't forget your | |
171 | VM/CT data) | |
172 | ||
173 | ||
174 | Remove Cluster configuration | |
175 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
176 | ||
177 | The recommended way is to reinstall the node after you removed it from | |
178 | your cluster. This makes sure that all secret cluster/ssh keys and any | |
179 | shared configuration data is destroyed. | |
180 | ||
181 | In some cases, you might prefer to put a node back to local mode | |
182 | without reinstall, which is described here: | |
183 | ||
184 | * stop the cluster file system in `/etc/pve/` | |
185 | ||
186 | # systemctl stop pve-cluster | |
187 | ||
188 | * start it again but forcing local mode | |
189 | ||
190 | # pmxcfs -l | |
191 | ||
192 | * remove the cluster configuration | |
193 | ||
194 | # rm /etc/pve/cluster.conf | |
195 | # rm /etc/cluster/cluster.conf | |
196 | # rm /var/lib/pve-cluster/corosync.authkey | |
197 | ||
198 | * stop the cluster file system again | |
199 | ||
200 | # systemctl stop pve-cluster | |
201 | ||
202 | * restart PVE services (or reboot) | |
203 | ||
204 | # systemctl start pve-cluster | |
205 | # systemctl restart pvedaemon | |
206 | # systemctl restart pveproxy | |
207 | # systemctl restart pvestatd | |
208 | ||
209 | ||
210 | ifdef::manvolnum[] | |
211 | include::pve-copyright.adoc[] | |
212 | endif::manvolnum[] |