logo

Manual Pages


Table of Contents

NAME

na_savecore - save a core dump

SYNOPSIS

savecore [ -i | -l | -s ]

savecore [ -f | -k | -w ] [ core id ]

DESCRIPTION

savecore is meant to be called near the end of the initialization file /etc/rc. Its function is to save the core dump of the system (assuming one was made) and to write the panic string to /etc/messages. savecore saves the compressed core dump file as core.sysid.YYYY-MM-DD.HH_MM_SS[.n].nz in the coredump directory (/etc/crash). A smaller, minimal core file, mini-core.sysid.YYYY-MM-DD.HH_MM_SS[.n] is also stored in the coredump directory. The timestamp embedded in the core and mini-core name is the time of the panic, in UTC.

Before savecore writes out a core image, it reads a number from the file /etc/crash/minfree. The minfree file defines the number of free kilobytes that should be left in the root filesystem after savecore runs. It should contain only a number followed by a newline. If the amount of space left after saving a core would be less than this number, savecore will abort, and not save any more cores. If the minfree file does not exist, savecore will save cores until it runs out of space in the root filesystem.

savecore is done in parallel with enabling services such as NFS. If a panic occurs prior to the completion of the savecore then the filer will do a synchronous savecore once it reboots.

When run with no arguments, savecore will save all cores that do not require the -f flag to be saved. Then, a mini-core will be generated for each saved core. If a core id is given, only the specified core will be considered.

A core id is a large integer and only applies to unsaved cores. They can be found by running savecore -l.

OPTION

-f
If savecore fails to save an unsaved core too many times, it will stop trying. This flag tells savecore to ignore the previous attempts, and try again. This attempt will be made synchronously.

-i
Displays information about the type of coredump that would be performed, if the system were to panic right now. If unsaved cores are present on the disks, this information will be included in the output.

-k
Invalidates the special core area that resides on each of the disks. Typically this is used when a savecore command cannot complete due to an error such as a disk read error.

-l
Lists all of the saved core files currently in the coredump directory, along with the panic string, panic time, and OS version. This information, along with a core id, is printed for unsaved cores.

-s
Displays the progress of the current savecore operation. If no save is in progress, but an unsaved core is present, a warning will be printed if the unsaved core cannot fit in the root filesystem.

-w
Starts savecore in a synchronous mode, which blocks other processing until the save has completed. Periodic progress updates will be displayed on the console.

FILES

/etc/crash/core.*.nz
saved core files
/etc/crash/mini-core.*
minimal core files
/etc/crash/minfree
free KB in FS to maintain after savecore

CLUSTER CONSIDERATIONS

If the live filer in a cluster has the savecore command in its /etc/rc file, it can save the cores created by its partner when its partner crashes. The cores are saved to the partner's coredump directory.

SEE ALSO

na_partner(1), na_rc(5).


Table of Contents