ECCLOGGER(1M)
Model 350 only
NAME
ecclogger − checks ECC memory for errors
eccscrub − scrubs soft errors from ECC memory
SYNOPSIS
ecclogger [logsize]
eccscrub
DESCRIPTION
ecclogger checks ECC (error-checking and correcting) memory, installed in an HP9000 Model 350 computer system, for errors which have been corrected. When ecclogger finds an error it writes a record of the error to /etc/ecclog and invokes eccscrub to insure that all errors have been cleared (only one error per ECC memory board is logged).
The optional parameter logsize specifies the number of log entries which are allowed. Hewlett-Packard recommends that the default value of 100 be used for logsize; if 100 errors occur on a system the local HP Sales and Service office should be notified, so that the /etc/ecclog can be evaluated (a system with less than 100 ecclog entries is generally not a concern). When the log actually contains this many entries a message is printed (mailed to root, when invoked from root’s crontab) to indicate that /etc/ecclog is full.
ecclogger should be invoked by root’s crontab(1). In order for this to occur an entry must be made in the crontab(1M) entry for root. The recommended frequency for ecclogger is once per hour. (eccscrub is automatically invoked by ecclogger whenever an error is corrected.)
eccscrub is a utility that will perform a read-modify-write function on memory. This function will correct any single-bit soft (not a ’stuck’ bit) errors that may be in a memory cell. It may be desirable to have eccscrub invoked by cron(1) once per day.
In order to achieve the recommended frequencies, the following two entries need to be added to the crontab(1M) entry for root: 0 * * * * exec /etc/ecclogger
0 0 * * * exec /etc/eccscrub
Below is a typical entry for a failure, in /etc/ecclog: 870911084132 0xFF55CF40 0x70
(Note: Do not write to /etc/ecclog. This will prevent ecclogger from writing to /etc/ecclog.)
The first field in the error entry is a date/time stamp [yymmddhhmmss] that indicates when the error was logged. The second field is the memory location in hexadecimal. The final field is the error syndrome byte. The syndrome byte is decoded below (primarily useful to Hewlett-Packard for repairing the ECC boards): SYNDROME ERROR LOCATION SYNDROME ERROR LOCATION
0x01 check bit 0 0x31 data bit 14
0x02 check bit 1 0x34 data bit 15
0x04 check bit 2 0x40 check bit 6
0x08 check bit 3 0x4A data bit 1
0x0B data bit 17 0x4F data bit 0
0x0E data bit 16 0x52 data bit 2
0x10 check bit 4 0x54 data bit 3
0x13 data bit 18 0x57 data bit 4
0x15 data bit 19 0x58 data bit 5
0x16 data bit 20 0x5B data bit 6
0x19 data bit 21 0x5D data bit 7
0x1A data bit 22 0x62 data bit 24
0x1C data bit 23 0x64 data bit 25
0x20 check bit 5 0x67 data bit 26
0x23 data bit 8 0x68 data bit 27
0x25 data bit 9 0x6B data bit 28
0x26 data bit 10 0x6D data bit 29
0x29 data bit 11 0x70 data bit 30
0x2A data bit 12 0x75 data bit 31
0x2C data bit 13
FILES
/etc/ecclog
SEE ALSO
Hewlett-Packard Company — June 06, 1988