Management Solaris Facility (SMF) - Workshop
Ganesh Hiregoudar Renaud Manus OP/N1 RPE Approachability
Sun Microsystems
Thanks!
We would like to thanks the following engineers who participated in writing and delivering this SMF workshop. > Jarod Nash, > Jason Banham, > Lee Brooks, > Rhodri Davies, > Phill Hughes
Terminology
Service
> Object (application, software state of a device, set of
other services) that can be managed and observed.
Instance
> Child of service object.
FMRI (Fault Managed Resource Identifier)
> 3 components:
svc: /system/system-log :default
Restarter
Service Type
Name
Instance
> Service responsible for restarting service(s)
Terminology (cont.)
Dependency
> Formal description of the other services that are required
to start a service
Milestone
> Predefined set of capabilities for a set of services.
Contract
> Mechanism within the kernel for restarters to stay
informed about the service instances they start.
Terminology (cont.)
Manifest
> Description and initial configuration file for a service.
Delivered with the product in the form of an XML file.
Repository
> Configuration database for all services. Allows setting to
remain persistent across reboots.
Snapshot
> Historical data about each service instance in the
repository for administrative inspection and rollback.
Key files and directories
/var/svc/manifest/* /lib/svc/method/* /var/svc/log/* /var/svc/profile/* /etc/svc/repository.db
> Global configuration database
/lib/svc/see/global.db
> Minimal configuration database
New filesystems
/etc/svc/volatile
> Contains transient data, eg. lock files, init state and some
log files. Allocated from swap.
/system/contracts
> Primary interface into contract subsystem, to allow
service instances to be restarted. Indirectly related to SMF.
SMF commands
svcs : report the status of service instances svcadm : manipulate services instances
> eg. enable, disable, restart, ...
svccfg : manipulate the repository
> eg. import, export, configure, ... services
svcprop : display properties for a given service instance inetadm : report and configure inetd based services inetconv : convert and import services from inetd.conf
Commands: svcs(1)
List enabled or all (-a) instances, sorted by state, time Explanations for errors/states (-x) Show dependencies (-d) and dependents (-D) Show member processes (-p), additional details (-v/-l)
$ svcs STATE STIME .... online 18:18:30 online 18:18:29 .... $ svcs -p sendmail STATE STIME online 18:18:29 18:18:29 18:18:29 FMRI svc:/network/http:apache2 svc:/network/smtp:sendmail FMRI svc:/network/smtp:sendmail 100180 sendmail 100181 sendmail
$ svcs -d sendmail STATE STIME FMRI online 18:17:44 svc:/system/identity:domain online 18:17:52 svc:/network/service:default ....
Commands: svcs(1)
You can use a pattern which uses globbing rules to build regular expressions svc:/ is automatically added to the pattern if omitted Examples:
$ svcs '*print*' $ svcs cron $ svcs dns/client
Commands: svcadm(1M)
Enable, disable, refresh, restart service instances Mark in special states (maintenance) Synchronously wait for changes (-s)
$ grep lianep /etc/user_attr lianep::::auths=solaris.smf.modify,solaris.smf.manage $ svcs apache2 STATE STIME FMRI ? svc:/network/http:apache2 $ # create /etc/apache2/httpd.conf $ svcadm enable apache2 STATE STIME FMRI online 19:19:01 svc:/network/http:apache2 $ # edit /etc/apache2/httpd.conf $ svcadm refresh apache2 $ svcs apache2 STATE STIME FMRI online 19:19:33 svc:/network/http:apache2 $ svcadm disable apache2 $ svcs apache2 STATE STIME FMRI disabled 19:20:07 svc:/network/http:apache2 See also http://blogs.sun.com/roller/page/sch/20040923
Commands: svcadm(1M) actions
enable: allow start once dependencies are satisfied disable: stop service and do not allow it to start again
> -t: enable/disable until the system is rebooted > -s: enable/disable synchronously (wait for it...)
refresh: reload service configuration and run the refresh method (if any) restart: stop the service, then allow it to start once its dependencies are satisfied (no configuration change made) clear: mark service as repaired
Commands: svcadm(1M) milestone
Milestone: A service which specifies a collection of dependencies which declare a specific state of system-readiness
> init S, boot -s
and friends still work
Major milestones, which are analogous to system runlevels can be reached directly
> from boot (-m milestone=), > the standard init invocations, > or via svcadm: milestone/single-user,
milestone/multi-user, milestone/multi-user-server
See also http://blogs.sun.com/roller/page/lianep/20050204
Commands: svcprop(1)
List properties of services and instances Fetch in convenient forms for scripting View running or current props (-c), uncomposed (-C)
$ svcprop network/http:apache2 ... physical/entities fmri svc:/network/physical:default physical/grouping astring optional_all physical/restart_on astring error physical/type astring service start/exec astring /lib/svc/method/http-apache2\ start start/timeout_seconds count 60 start/type astring method stop/exec astring /lib/svc/method/http-apache2\ stop stop/timeout_seconds count 60 stop/type astring method restarter/auxiliary_state astring none restarter/next_state astring none restarter/state astring disabled restarter/state_timestamp time 1102030556.737590000 $ svcprop -p enabled network/http:apache2 false
Administration exercise
Write a script which:
> > > >
prints the start/exec property for sendmail enables sendmail waits for the service to start prints the service state
Administration exercise
Write a script which:
> > > >
prints the start/exec property for sendmail enables sendmail waits for the service to start prints the service state
#!/bin/ksh /usr/bin/svcprop -p start/exec svc:/network/smtp:sendmail /usr/sbin/svcadm enable -s svc:/network/smtp:sendmail if [ $? -ne 0 ]; then echo sendmail failed to start fi /usr/bin/svcprop -p restarter/state svc:/network/smtp:sendmail
Commands: inetadm(1M)
List services managed by inetd View (-l) and modify (-m) inetd-specific properties
$ inetadm ... enabled enabled disabled disabled enabled disabled enabled online online disabled disabled online disabled online svc:/network/ftp:default svc:/network/finger:default svc:/network/login:eklogin svc:/network/login:klogin svc:/network/login:rlogin svc:/network/rexec:default svc:/network/shell:default
$ inetadm -l ftp SCOPE NAME=VALUE name="ftp" endpoint_type="stream" proto="tcp6" isrpc=FALSE wait=FALSE exec="/usr/sbin/in.ftpd -a" user="root" ... default tcp_wrappers=FALSE
svccfg(1M)
svccfg manipulates the repository
Uses sub-commands to perform actions
Common sub-commands
select <fmri> list listprop setprop delete validate <file> import <file> unselect : select a service / instance : show children of the selected service : display the properties for a service : change a property value for a service : delete a service / instance : validate an XML manifest file : import a manifest file into repository : navigate to parent of current selection
Commands: svccfg(1M)
Import, export manifests; apply, extract profiles Interactive mode for modifying properties
$ svccfg -v import /var/svc/manifest/network/http-apache2.xml svccfg: Refreshed network/http:/apache2 svccfg: Successful import. $ svccfg svc:> select network/http:apache2 svc:/network/http:apache2> listprop ... general framework general/enabled boolean false ... start method start/exec astring "/lib/svc/method/http-apache2 start" start/timeout_seconds count 60 start/type astring method svc:/network/http:apache> editprop [$EDITOR launches, allows direct editing of properties] svc:/network/http:apache2> exit $ svcadm refresh apache2 $ svcadm restart apache2 # read changed config # restart with changed config
svccfg(1M): example (cont.)
To change a property value:
> Use the listprop sub-command to view properties > Use the setprop sub-command to set values
Example:
Increase the file descriptor limit for lpsched
# svccfg svc:> select application/print/server svc:/...> listprop svc:/...> setprop lpsched/fd_limit = 8192 svc:/...> quit # svcadm refresh print/server:default # svcadm restart print/server:default
svcs - example
Checking dependencies (NFS server)
# svcs -d svc:/network/nfs/server:default STATE STIME FMRI online Nov_15 svc:/network/loopback:default online Nov_15 svc:/network/physical:default online Nov_15 svc:/network/rpc/bind:default online Nov_15 svc:/network/rpc/keyserv:default online Nov_15 svc:/system/filesystem/local:default online Nov_15 svc:/network/rpc/gss:ticotsord online Nov_15 svc:/network/nfs/mapid:default online 10:32:25 svc:/network/nfs/nlockmgr:default
inetadm - examples
Changing properties of an inetd service (in.ftpd)
# inetadm -l svc:/network/ftp:default SCOPE NAME=VALUE name="ftp" endpoint_type="stream" proto="tcp6" isrpc=FALSE wait=FALSE exec="/usr/sbin/in.ftpd -a" user="root" . . . # inetadm -m svc:/network/ftp:default exec="/usr/sbin/in.ftpd -a -l" # inetadm -l svc:/network/ftp:default | grep exec exec="/usr/sbin/in.ftpd -a -l"
Misc - examples
Milestone transition: None to Single User
ok boot -m milestone=none
...
# svcs '*milestone*' STATE STIME disabled 13:12:49 disabled 13:12:49 disabled 13:13:04 [edited for brevity]
FMRI svc:/milestone/multi-user:default svc:/milestone/multi-user-server:default svc:/milestone/single-user:default
# svcadm milestone svc:/milestone/single-user:default # svcs | grep '\*' offline* 13:32:26 svc:/milestone/single-user:default offline* 13:32:28 svc:/system/device/local:default
Advanced concepts
Service instance states Repository Manifest Snapshots Profiles Demo
Service instance states
Uninitialised The service instance is not running and the configuration data has not been read. Usually seen if you have booted to milestone=none OfflineThe service instance is not running although the configuration data has been read. This is usually the result of a dependency that has not been met, or if there is an error in the start method. Online Disabled The service instance is running. All of its dependencies have been met. A service instance is not running. This may be the default state for a service when it is first imported into the repository, or an administrator may have marked the service as disabled. It will require operator assistance to move out of this state. A limited set of failures (usually dependencies) may cause the service instance to function in a limited capacity, eg: if an inetd service can work with IPv4 and IPv6 but the latter is not configured, then only IPv4 will be in use and thus the service may be in a degraded state.
Degraded
Maintenance The service instance is unavailable due to an error. There are many reasons why, ranging from unsatisfied dependencies and failed start methods, to more complicated reasons.
Repository
The repository is THE source for all known services on the system Modifications to service configuration and state are stored in the repository Services live in the repository! The repository is an SQLite database
Manifest
DTD is the reference to write manifest
> /usr/share/lib/xml/dtd/service_bundle.dtd.1
Explore /var/svc/manifest for similar services
> system/utmp is a simple standalone deamon > system/coreadm is a simple configuration service > network/telnet is an inet-managed deamon
Initial inet service manifests can be created easily by invoking inetconv -i <file>
Manifest (order is important)
Service bundle Service name Default instance Single instance Dependencies Exec. Methods Property groups Stability level Template
Manifest (cont.)
Basic elements
> > > > > > > >
Name of the service Number of instances Start, stop, [refresh] methods Dependencies Property groups Service model Fault handling section Documentation template
Advanced elements
Service relationships
Milestone multi-user-server DEPENDENT DEPENDENCY
svc:/network/nfs/server
Dependency
The other services that a service relies upon in order to start A mechanism for a service to declare itself as a dependency of another service
Dependent
DEPENDENCIES
svc:/system/filesystem/local
Colour Key:
Implicit Relationship Explicit Relationship
svc:/network/rpc/gss svc:/network/rpc/keyserv svc:/network/rpc/bind
Fig.1: Relationship diagram for svc:/network/nfs/server
Service relationships
> Restart relationships defined by dependency groups > grouping
require_all : all services are running require_any: at least one service is running optional_all: all services running, disabled or maintenance exclude_all : all services disabled, maint. or absent
> restart_on
none : required only for startup error : restart if dependency fails due to hw or sw error restart : restart if dependency restart to any reason refresh : restart if dependency restarts or is refreshed
Manifest: dependents
A service manifest can specify which services are dependent upon it. It is forbidden to modify the manifest of other services For predictable startup behaviour, the service manifest should specify one of the major milestones as a dependent.
<dependent name='nfs-client_multi-user' grouping='optional_all' restart_on='none'> <service_fmri value='svc:/milestone/multi-user' /> </dependent>
Development:
utmpd(1M) example
<service name='system/utmp' type='service' version='1'> <create_default_instance enabled='true' /> <single_instance /> <dependency name='milestone' grouping='require_all' restart_on='none' type='service'> <service_fmri value='svc:/milestone/sysconfig'/> </dependency> <dependent name='utmpd_multi-user' grouping='optional_all' restart_on='none'> <service_fmri value='svc:/milestone/multi-user'/> </dependent> <exec_method type='method' name='start' exec='/lib/svc/method/svc-utmpd' timeout='60' /> <exec_method type='method' name='stop' exec=':kill' timeout='60' /> <stability value='Unstable' /> <template> <common_name><loctext xml:lang='C'> utmpx monitoring </loctext></common_name> <documentation> <manpage title='utmpd' section='1M' manpath='/usr/share/man' /> </documentation> </template> </service>
Snapshots
A snapshot is a copy of the configuration information for a service instance Can't be used for 'cloning' Up to 5 levels of snapshot
> initial : initial conf. delivered by a package > last-import: conf. of an instance when last imported by
svccfg(1M) > running : the running conf. of the instance > start : conf. captured during a successful transition to online state > previous : conf. captured when an undo operation is performed
Snapshots (cont.)
Snapshots taken at the instance level Most configuration settings at service level Use svccfg(1M) to select an instance
> listsnap : list available snapshots > selectsnap : select one of the snapshots > revert [snap] : revert back to a given snapshot
The changed property values can be made active via 'svcadm refresh'
Snapshots: example
[...] svc:/network/smtp:sendmail> listsnap initial running start svc:/network/smtp:sendmail> selectsnap running [running]svc:/network/smtp:sendmail> listprop config-file/* config-file/entities fmri file://localhost/etc/mial/sendmail.cf config-file/grouping astring require_all config-file/restart_on astring refresh config-file/type astring path [running]svc:/network/smtp:sendmail> selectsnap start [start]svc:/network/smtp:sendmail> listprop config-file/* config-file/entities fmri file://localhost/etc/mail/sendmail.cf config-file/grouping astring require_all config-file/restart_on astring refresh config-file/type astring path [start]svc:/network/smtp:sendmail> revert svc:/network/smtp:sendmail> listsnap initial running start previous # svcadm refresh svc:/network/smtp:sendmail
Profiles
A profile is a description of the services that are to be used on a system
> Applied in order : generic, platform, site > Profiles may include sub-profiles
Enable/Disable services for your site
> eg. jumpstart, JASS
JumpStart Architecture and Security Scripts
Profile location : /var/svc/profile/* Never modify existing profiles The site.xml file is for local customizations, which you must create
Profiles (cont.)
To create your own site.xml file:
svccfg:> extract > /var/tmp/site.xml (modify file with your changes) # cp /var/tmp/site.xml /var/svc/profile # init 6
You can also svccfg apply the site.xml for immediate effect Be aware, profiles are only processed once
Debug & Lab
Troubleshooting techniques Lab : Writing your own service Lab exercises
Troubleshooting
Service failures printed to console, syslog Always start with svcs -x; http://sun.com/msg
> svcs -x
will display the service logfile, if it exists:
> /var/svc/log > /etc/svc/volatile
See service start messages with boot For a system that hangs during boot:
> boot -m verbose,milestone=none
-m verbose
> log in at prompt
> svcadm milestone all
> Watch system progress with svcs
Troubleshooting (cont.)
To truss a service starting up
> Truss the restarter (svc.startd/inetd)
Or
> Make a note of the start method > Use svccfg to modify the start method > When complete, restore the original start method
# svccfg svc:> select application/print/rfc1179 svc:...1179> listprop inetd_start/exec inetd_start/exec astring /usr/lib/print/in.lpd svc:...1179> setprop inetd_start/exec = /usr/bin/truss -ealfo /tmp/in.lpd.truss /usr/lib/print/in.lpd svc:...1179> end # svcadm refresh svc:/application/print/rfc1179:default
Troubleshooting (cont.)
Maintenance mode
> svcs -xv or svcs -l to see why the service failed > svcadm refresh <fmri> if repository has been changed > svcadm clear <fmri> once fault rectified
Have changes been made to repository?
> How services many are affected? > What snapshots do exist? > Can you revert back to a previous snapshot?
Troubleshooting (cont.)
If the current repository is beyond repair, try to restore from an earlier repository or reseed:
/lib/svc/bin/restore_repository
Will the system boot in verbose mode?
> ok boot -m verbose
Will the system boot with no service?
> ok boot -m milestone=none
Lab Unable to resolve webcache
Ru n th e te s t1 .b re a km e s crip t
# nslookup webcache ;; connection timed out; no servers could be reached # svcs svc:/network/dns/client:default offline 17:34:25 svc:/network/dns/client:default
Lab Unable to resolve webcache
# svcs -xv svc:/network/dns/client:default svc:/network/dns/client:default (DNS resolver) State: offline since Fri Feb 04 18:10:24 2005 Reason: Dependency file://localhost/etc/resolv.conf is absent. See: http://sun.com/msg/SMF-8000-E2 See: man -M /usr/share/man -s 3RESOLV resolver See: /var/svc/log/network-dns-client:default.log Impact: This service is not running.
In this case it clearly tells us that /etc/resolv.conf is missing
# svcadm refresh svc:/network/dns/client:default
Lab sendmail won't start
Ru n t h e t e s t 2 .b re a k m e s c rip t
svc.startd[7]: network/smtp:sendmail failed # svcs svc:/network/smtp:sendmail STATE STIME FMRI maintenance 16:18:03 svc:/network/smtp:sendmail
Lab sendmail won't start
# svcs -xv svc:/network/smtp:sendmail svc:/network/smtp:sendmail (sendmail SMTP mail transfer agent) State: maintenance since Mon Nov 22 16:18:03 2004 Reason: Start method failed repeatedly, last died on Killed (9). See: http://sun.com/msg/SMF-8000-KS See: man -M /usr/man -s 1M sendmail See: /var/svc/log/network-smtp:sendmail.log Impact: This service is not running.
# tail /var/svc/log/network-smtp:sendmail.log [ Nov 22 16:17:53 executing start method ("/lib/svc/method/smtp-sendmail start") ] [ Nov 22 16:18:03 Method or service exit timed out. Killing contract 122 ] [ Nov 22 16:18:03 Method "start" failed due to signal Killed ]
Lab sendmail won't start
Knowing the start method timed out we can examine the repository for this service:
# svcprop -p start svc:/network/smtp:sendmail start/exec astring /lib/svc/method/smtp-sendmail\ start start/type astring method start/timeout_seconds count 5
Having extracted this value we can compare it with the value in the manifest:
# cd /var/svc/manifest/network # more smtp-sendmail.xml <?xml version="1.0"?> ... <exec_method type='method' name='start' exec='/lib/svc/method/smtp-sendmail start' timeout_seconds='120' />
Lab sendmail won't start
1) Use svccfg to set the correct start method timeout: # svccfg svc:> select svc:/network/smtp:sendmail svc:/network/smtp:sendmail> setprop start/timeout_seconds = 120 svc:/network/smtp:sendmail> quit # svcadm refresh svc:/network/smtp:sendmail # svcadm clear svc:/network/smtp:sendmail # svcs svc:/network/smtp:sendmail STATE online STIME FMRI 17:03:48 svc:/network/smtp:sendmail
Lab sendmail won't start
2) Use snapshots to set the correct start method timeout: # svccfg svc:> select svc:/network/smtp:sendmail svc:/network/smtp:sendmail> selectsnap initial [initial]svc:/network/smtp:sendmail> revert svc:/network/smtp:sendmail> quit # svcadm refresh svc:/network/smtp:sendmail # svcadm clear svc:/network/smtp:sendmail # svcs svc:/network/smtp:sendmail STATE online STIME FMRI 17:03:48 svc:/network/smtp:sendmail
Lab keyserv won't start
Ru n th e te s t3 .b re a km e s crip t
svc.startd[7]: svc:/network/nis/client:default: Method "/lib/svc/method/yp" failed with exit status 96. svc.startd[7]: svc:/network/rpc/keyserv:default: Method "/usr/sbin/keyserv" failed with exit status 96. # svcs svc:/network/nis/client:default STATE STIME FMRI maintenance 10:52:40 svc:/network/nis/client:default # svcs svc:/network/rpc/keyserv:default STATE STIME FMRI maintenance 10:52:40 svc:/network/rpc/keyserv:default
Lab keyserv won't start
# svcs -xv network/rpc/keyserv:default svc:/network/rpc/keyserv:default (RPC encryption key storage) State: maintenance since Wed Feb 09 17:20:46 2005 Reason: Start method exited with SMF_EXIT_ERR_CONFIG. See: http://sun.com/msg/SMF-8000-KS See: keyserv(1M) See: /var/svc/log/network-rpc-keyserv:default.log Impact: This service is not running. # svcs -xv network/nis/client:default svc:/network/nis/client:default (NIS (YP) client) State: maintenance since Wed Feb 09 17:20:46 2005 Reason: Start method exited with $SMF_EXIT_ERR_CONFIG. See: http://sun.com/msg/SMF-8000-KS See: ypstart(1M) See: /var/svc/log/network-nis-client:default.log Impact: This service is not running.
Lab keyserv won't start
# cd /var/svc/log # tail network-rpc-keyserv:default.log [ Feb [ Feb [ Feb [ Feb 9 17:17:19 Stopping because service disabled. ] 9 17:17:20 Executing stop method (:kill) ] 9 17:20:45 Executing start method (/usr/sbin/keyserv) ] 9 17:20:46 Method start exited with status 96 ]
# tail network-nis-client:default.log [ Feb [ Feb [ Feb [ Feb 9 17:17:19 Stopping because service disabled. ] 9 17:17:20 Executing stop method (:kill) ] 9 17:20:45 Executing start method (/lib/svc/method/yp) ] 9 17:20:46 Method start exited with status 96 ]
/lib/svc/method/yp: domainname not set
# ls -l /etc/defaultdomain /etc/defaultdomain: No such file or directory
Lab keyserv won't start
# cat > /etc/defaultdomain louvre.France.Sun.COM ^D # domainname `cat /etc/defaultdomain` # svcadm clear nis/client:default # svcs nis/client:default STATE online STIME FMRI 9:59:21 svc:/network/nis/client:default
# svcadm clear network/rpc/keyserv:default # svcs network/rpc/keyserv:default STATE online STIME FMRI 10:01:46 svc:/network/rpc/keyserv:default
Lab customer ndd script
Ru n th e te s t4 .b re a km e s crip t
Feb 10 12:25:21 labhost svc.startd[7]: network/ndd-nettune:default failed repeatedly
# svcs svc:/network/ndd-nettune:default STATE STIME FMRI maintenance 16:02:26 svc:/network/ndd-nettune:default
Lab customer ndd script
# svcs -xv network/ndd-nettune:default svc:/network/ndd-nettune:default (ndd network tuning) State: maintenance since Thu Feb 10 12:25:21 2005 Reason: Restarting too quickly. See: http://sun.com/msg/SMF-8000-L5 See: man -M /usr/share/man -s 1M ndd See: /var/svc/log/network-ndd-nettune:default.log Impact: This service is not running. # tail /var/svc/log/network-ndd-nettune:default.log [ Feb 11 15:12:14 Executing start method ("/lib/svc/method/ndd-nettune") ] /sbin/sh: /lib/svc/method/ndd-nettune: cannot execute [ Feb 11 15:12:14 Stopping because all processes in service exited. ] [ Feb 11 15:12:14 Executing start method ("/lib/svc/method/ndd-nettune") ] /sbin/sh: /lib/svc/method/ndd-nettune: cannot execute [ Feb 11 15:12:14 Stopping because all processes in service exited. ] [ Feb 11 15:12:14 Executing start method ("/lib/svc/method/ndd-nettune") ] /sbin/sh: /lib/svc/method/ndd-nettune: cannot execute [ Feb 11 15:12:14 Stopping because all processes in service exited. ] [ Feb 11 15:12:14 Restarting too quickly, changing state to maintenance ]
Lab customer ndd script
# ls -l /lib/svc/method/ndd-nettune -rw-r--r-1 root root 477 Feb 11 15:12 /lib/svc/method/ndd-nettune # chmod 755 /lib/svc/method/ndd-nettune # svcadm clear network/ndd-nettune:default Feb 11 15:19:51 labhost svc.startd[7]: network/ndd-nettune:default failed repeatedly # more /var/svc/log/network-ndd-nettune:default.log [ Feb 11 15:19:48 Leaving maintenance because clear requested. ] [ Feb 11 15:19:49 Enabled. ] [ Feb 11 15:19:49 Executing start method ("/lib/svc/method/ndd-nettune") ] [ Feb 11 15:19:49 Stopping because all processes in service exited. ] [ Feb 11 15:19:49 Executing start method ("/lib/svc/method/ndd-nettune") ] [ Feb 11 15:19:49 Stopping because all processes in service exited. ] [ Feb 11 15:19:49 Executing start method ("/lib/svc/method/ndd-nettune") ] [ Feb 11 15:19:49 Stopping because all processes in service exited. ] [ Feb 11 15:19:51 Restarting too quickly, changing state to maintenance ]
Lab customer ndd script
# cd /lib/svc/method # file ndd-nettune ndd-nettune: executable /sbin/sh script # more /lib/svc/method/ndd-nettune #!/sbin/sh # # ident "@(#)ndd-nettune.xml 1.0 04/09/21 SMI"
. /lib/svc/share/smf_include.sh . /lib/svc/share/net_include.sh # Make sure that the libraries essential to this stage of booting can be found. LD_LIBRARY_PATH=/lib; export LD_LIBRARY_PATH echo "ndd ran" >> /tmp/smf.out /usr/sbin/ndd -set /dev/tcp tcp_recv_hiwat 16384 /usr/sbin/ndd -set /dev/tcp tcp_xmit_hiwat 16384 # Reset the library path now that we are past the critical stage unset LD_LIBRARY_PATH
Lab customer ndd script
A transient script should run once then exit and never be restarted. In the log file you can see that the start method is being run repeatedly. This suggests the service has been configured incorrectly, so you should look at the repository: # svccfg svc:> select network/ndd-nettune svc:/network/ndd-nettune> listprop startd/* startd/duration astring child
Here you see this service has been mis-configured as a child service, so when the process died svc.startd attempted to restart it. transient so: svc:/network/ndd-nettune> setprop startd/duration = transient svc:/network/ndd-nettune> quit # svcadm refresh svc:/network/ndd-nettune:default # svcadm clear svc:/network/ndd-nettune:default STATE online STIME FMRI 15:20:18 svc:/network/ndd-nettune:default We know it should be
Service Lifecycle
Determine service suitability Write service methods Write service manifest Import manifest Test/modify Package manifest and methods Install service Service startup/runtime Configure/modify service Remove service
Method creation
Start with /etc/init.d script Include /lib/svc/share/smf_include.sh
> SMF_EXIT_OK > SMF_EXIT_ERR_FATAL > SMF_EXIT_ERR_CONFIG
Fail with a non-0 exit code, and emit a message to stdout or stderr start and stop required, refresh optional :kill [signal] and :true Don't exit until service is completely started remove script from /etc/rc?.d locations and /etc/init.d
Write your own service
Tip : How to check the manifest? > Use xmllint(1) or Mozilla to parse the manifest and check for errors
> Example:
# xmllint mysvc.xml mysvc.xml:71: parser error : Opening and ending tag mismatch: exec_method line 36 and service </service> ^ mysvc.xml:73: parser error : expected '>' </service_bundle> ^ mysvc.xml:74: parser error : Premature end of data in tag service_bundle line 10
Write your own service
little d : A simple daemon that listens on port 13567
> Requires a configuration file: /var/tmp/littled.conf
You can interact with littled thus:
$ telnet localhost 13567 SMF TOI littled daemon/server version=0.8 Daemon Started: Wed Feb 09 11:07:44 2005 littled> help SMF TOI littled daemon/server version=0.8 Available commands: ...