| 1 |
The events/ directory contains event scripts used by CTDB. Event |
| 2 |
scripts are triggered on certain events, such as startup, monitoring |
| 3 |
or public IP allocation. Scripts may be specific to services, |
| 4 |
networking or internal CTDB operations. |
| 5 |
|
| 6 |
Scripts are divided into subdirectories for different CTDB components. |
| 7 |
Right now the only component is "legacy". |
| 8 |
|
| 9 |
All event scripts start with the prefix 'NN.' where N is a digit. The |
| 10 |
event scripts are run in sequence based on NN. Thus 10.interface will |
| 11 |
be run before 60.nfs. It is recommended to keep each NN unique. |
| 12 |
However, scripts with the same NN prefix will be executed in |
| 13 |
alphanumeric sort order. |
| 14 |
|
| 15 |
As a special case, any eventscript that ends with a '~' character will be |
| 16 |
ignored since this is a common postfix that some editors will append to |
| 17 |
older versions of a file. Similarly, any eventscript with multiple '.'s |
| 18 |
will be ignored as package managers can create copies with additional |
| 19 |
suffix starting with '.' (e.g. .rpmnew, .dpkg-dist). |
| 20 |
|
| 21 |
Only executable event scripts are run by CTDB. Any event script that |
| 22 |
does not have execute permission is ignored. |
| 23 |
|
| 24 |
The eventscripts are called with varying number of arguments. The |
| 25 |
first argument is the event name and the rest of the arguments depend |
| 26 |
on the event name. |
| 27 |
|
| 28 |
Event scripts must return 0 for success and non-zero for failure. |
| 29 |
|
| 30 |
Output of event scripts is logged. On failure the output of the |
| 31 |
failing event script is included in the output of "ctdb scriptstatus". |
| 32 |
|
| 33 |
The following events are supported (with arguments shown): |
| 34 |
|
| 35 |
init |
| 36 |
|
| 37 |
This event is triggered once when CTDB is starting up. This |
| 38 |
event is used to do some basic cleanup and initialisation. |
| 39 |
|
| 40 |
During the "init" event CTDB is not listening on its Unix |
| 41 |
domain socket, so the "ctdb" CLI will not work. |
| 42 |
|
| 43 |
Failure of this event will cause CTDB to terminate. |
| 44 |
|
| 45 |
Example: 00.ctdb creates $CTDB_SCRIPT_VARDIR |
| 46 |
|
| 47 |
setup |
| 48 |
|
| 49 |
This event is triggered once, after the "init" event has |
| 50 |
completed. |
| 51 |
|
| 52 |
For this and any subsequent events the CTDB Unix domain socket |
| 53 |
is available, so the "ctdb" CLI will work. |
| 54 |
|
| 55 |
Failure of this event will cause CTDB to terminate. |
| 56 |
|
| 57 |
Example: 11.natgw checks that it has valid configuration |
| 58 |
|
| 59 |
startup |
| 60 |
|
| 61 |
This event is triggered after the "setup" event has completed |
| 62 |
and CTDB has finished its initial database recovery. |
| 63 |
|
| 64 |
This event starts all services that are managed by CTDB. Each |
| 65 |
service that is managed by CTDB should implement this event |
| 66 |
and use it to (re)start the service. |
| 67 |
|
| 68 |
If the "startup" event fails then CTDB will retry it until it |
| 69 |
succeeds. There is no limit on the number of retries. |
| 70 |
|
| 71 |
Example: 50.samba uses this event to start the Samba daemon. |
| 72 |
|
| 73 |
shutdown |
| 74 |
|
| 75 |
This event is triggered when CTDB is shutting down. |
| 76 |
|
| 77 |
This event shuts down all services that are managed by CTDB. |
| 78 |
Each service that is managed by CTDB should implement this |
| 79 |
event and use it to stop the service. |
| 80 |
|
| 81 |
Example: 50.samba uses this event to shut down the Samba |
| 82 |
daemon. |
| 83 |
|
| 84 |
monitor |
| 85 |
|
| 86 |
This event is run periodically. The interval between |
| 87 |
successive "monitor" events is configured using the |
| 88 |
MonitorInterval tunable, which defaults to 15 seconds. |
| 89 |
|
| 90 |
This event is triggered by CTDB to continuously monitor that |
| 91 |
all managed services are healthy. If all event scripts |
| 92 |
complete then the monitor event successfully then the node is |
| 93 |
marked HEALTHY. If any event script fails then no subsequent |
| 94 |
scripts will be run for that event and the node is marked |
| 95 |
UNHEALTHY. |
| 96 |
|
| 97 |
Each service that is managed by CTDB should implement this |
| 98 |
event and use it to monitor the service. |
| 99 |
|
| 100 |
Example: 10.interface checks that each configured interface |
| 101 |
for public IP addresses has a physical link established. |
| 102 |
|
| 103 |
startrecovery |
| 104 |
|
| 105 |
This event is triggered every time a database recovery process |
| 106 |
is started. |
| 107 |
|
| 108 |
This is rarely used. |
| 109 |
|
| 110 |
recovered |
| 111 |
|
| 112 |
This event is triggered every time a database recovery process |
| 113 |
is completed. |
| 114 |
|
| 115 |
This is rarely used. |
| 116 |
|
| 117 |
takeip <interface> <ip-address> <netmask-bits> |
| 118 |
|
| 119 |
This event is triggered for each public IP address taken by a |
| 120 |
node during IP address (re)assignment. Multiple "takeip" |
| 121 |
events can be run in parallel if multiple IP addresses are |
| 122 |
being assigned. |
| 123 |
|
| 124 |
Example: In 10.interface the "ip" command (from the Linux |
| 125 |
iproute2 package) is used to add the specified public IP |
| 126 |
address to the specified interface. The "ip" command can |
| 127 |
safely be run concurrently. However, the "iptables" command |
| 128 |
cannot be run concurrently so a wrapper is used to serialise |
| 129 |
runs using exclusive locking. |
| 130 |
|
| 131 |
If substantial work is required to reconfigure a service when |
| 132 |
a public IP address is taken over it can be better to defer |
| 133 |
service reconfiguration to the "ipreallocated" event, after |
| 134 |
all IP addresses have been assigned. |
| 135 |
|
| 136 |
Example: 60.nfs uses ctdb_service_set_reconfigure() to flag |
| 137 |
that public IP addresses have changed so that service |
| 138 |
reconfiguration will occur in the "ipreallocated" event. |
| 139 |
|
| 140 |
startipreallocate |
| 141 |
|
| 142 |
This event is triggered on all nodes before ip address is |
| 143 |
released on node. This can be used to perform action needed |
| 144 |
to complete before ip is given away to other node. |
| 145 |
|
| 146 |
Example: 60.nfs would use this event to put nfs-ganesha server |
| 147 |
on all nodes in grace period so that locks can be reclaimed |
| 148 |
safely in lock reclaim phase. |
| 149 |
|
| 150 |
releaseip <interface> <ip-address> <netmask-bits> |
| 151 |
|
| 152 |
This event is triggered for each public IP address released by |
| 153 |
a node during IP address (re)assignment. Multiple "releaseip" |
| 154 |
events can be run in parallel if multiple IP addresses are |
| 155 |
being unassigned. |
| 156 |
|
| 157 |
In all other regards, this event is analogous to the "takeip" |
| 158 |
event above. |
| 159 |
|
| 160 |
updateip <old-interface> <new-interface> <ip-address> <netmask-bits> |
| 161 |
|
| 162 |
This event is triggered for each public IP address moved |
| 163 |
between interfaces on a node during IP address (re)assignment. |
| 164 |
Multiple "updateip" events can be run in parallel if multiple |
| 165 |
IP addresses are being moved. |
| 166 |
|
| 167 |
This event is only used if multiple interfaces are capable of |
| 168 |
hosting an IP address, as specified in the public addresses |
| 169 |
configuration file. |
| 170 |
|
| 171 |
This event is similar to the "takeip" event above. |
| 172 |
|
| 173 |
ipreallocated |
| 174 |
|
| 175 |
This event is triggered on all nodes as the last step of |
| 176 |
public IP address (re)assignment. It is unconditionally |
| 177 |
triggered after any "releaseip", "takeip" and "updateip" |
| 178 |
events, even though these events may not run on some nodes if |
| 179 |
there are no relevant changes. That is, the "ipreallocated" |
| 180 |
event is triggered unconditionally, even on nodes where public |
| 181 |
IP addresses assignments have not changed. |
| 182 |
|
| 183 |
This event is used to reconfigure services. |
| 184 |
|
| 185 |
Since "ipreallocated" is always run, this allows |
| 186 |
reconfiguration to depend on the states of other nodes rather |
| 187 |
that just IP addresses. |
| 188 |
|
| 189 |
Example: 11.natgw recalculates the NAT gateway master and |
| 190 |
updates the relevant network configuration on each node if the |
| 191 |
NAT gateway master has changed. |
| 192 |
|
| 193 |
Additional notes for "takeip", "releaseip", "updateip", |
| 194 |
"ipreallocated": |
| 195 |
|
| 196 |
* Failure of any of these events causes IP allocation to be retried. |
| 197 |
|
| 198 |
* An event script can use ctdb_service_set_reconfigure() in "takeip", |
| 199 |
"releaseip" or "updateip" events to flag that its service needs to |
| 200 |
be reconfigured. The "ipreallocated" event can then use |
| 201 |
ctdb_service_needs_reconfigure() to test if there were public IPs |
| 202 |
changes to determine what type of reconfiguration (if any) is |
| 203 |
needed. |