English

Error: CLEAN returns: Write error (MSEEK), bailing out of CLNSRT

Solution or Workaround

Question:

What causes CLEAN to bail out with the following message?

Fatal error
Write error (MSEEK)
Bailing out of CLNSRT

Cause:

Not enough disk space to process the coverage.

Answer:

The situation may occur if the /tmp directory is full or there is insufficient
swap space. Check disk usage with 'df' on UNIX platforms and 'SHOW QUOTA' on
VAX platforms. If the disk is nearly full, backup infrequently used data and
delete the unnecessary files to free up space. Verify that no one is currently
using the system and then delete any temporary files in the /tmp directory or
in the current workspace (xx for UNIX and $T***.dat for the VAX). Check available swap space (use /etc/pstat -T on SunOS) and increase it if necessary (see Installing ARC/INFO, or the ARC/INFO System Administrator Guide for more information about swap space requirements). In general, swap space should be increased when greater demands are placed on a system (i.e., larger coverages, more users, multiple sessions of ARC, etc.).

Fine-tuning system parameters on VAX platforms is a bit more complicated.
Check for memory usage with the MONITOR CLUSTER command. If "%Memory In Use"
is high (90-100%), it is very likely that a "memory bottleneck" exists. Memory
bottlenecks also create a high page-fault rate (MONITOR PAGE). Lowering the
Authorize values for WSDEF, WSQUO, and WSEXT or limiting the number of
interactive processes on the system will improve performance. ESRI recommends
contacting DEC for advice about improving CPU and memory performance with
ARC/INFO. However, following three basic rules of thumb will significantly
improve system performance in a multiuser environment:

1. Run large, non-interactive jobs in batch queues during the night-time
hours.

2. Distribute disk access over as many disks as possible. Multiple users
accessing a single disk may cause I/O bottlenecks.

3. Limit system access to only as many interactive users as the system can
handle. Six users with limited system response will get less done than two
shifts of three users with an adequate system response time.

Notes:

Consult the "ARC/INFO System Administrator's Guide".