Attempt to access item beyond bounds of memory (& ulimit); "Failed to read reply
- Article Type: General
- Product: Aleph
- Product Version: 20
Description:
We have been experiencing numerous occurrences of this error -- there is no pattern as to a specific problem. We did not have these problems in test, but never had quite the volume of users that we are now experiencing. These errors are happening in both the pc and web servers.
What is the recommended ulimit size where memory is concerned? Here's what we have:
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 1048576
stack(kbytes) 32768
memory(kbytes) 524288
coredump(blocks) 0
nofiles(descriptors) 2000
I notice that on other of our library servers, ulimit for memory is set to unlimited.
[Later:]
We've now set our ulimits as follows and we continue to see errors::
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) unlimited
stack(kbytes) unlimited
memory(kbytes) unlimited
coredump(blocks) 0
nofiles(descriptors) 2000
The errors are much the same as previously, but, after making the changes are now entirely confined to the pc_server_main process.
get_buf_z108 72
pc_display_fill_html 35
form_fill_xml_record 12
www_f 9
get_buf_z30_adm 7
get_buf_z30_adm_by_iobuf 5
check_doc_line_single 3
io_word3_a 3
get_buf_z30 3
form_fill_xml_section 2
update_statistics 2
set_find_m 2
check_doc 2
io_word3 1
get_buf_z36_id 1
correct_z30 1
As a rule, these are not terribly disruptive, but do indicate that something is not quite right. Many of these I've been able to trigger myself as indicated previously -- move back and forth between patron list of loans and course list in the CIRC client after being idle for serveral minutes-- even without changing any information. Errors do not occur with the same activity when fewer users are logged in. Yesterday, errors began happening when 25 users were logged in. After 5:00, when most staff had gone home for the day, no errors occurred. We don't know quite where else to look at the moment.
Resolution:
We have made the necessary changes to the most likely causes of the problems and have rolled the revised programs into version 20 rep_change #2575.
Note: This problem is most common on AIX servers.
- Article last edited: 10/8/2013