Red Hat Bugzilla – Bug 1253851
Top displays high memory usage in a strange way
Last modified: 2018-04-10 09:41:03 EDT
Description of problem: A process that uses a considerable sum or far gets a strange line in top: 1865 jchem-p+ 20 0 41.722g 0.018t 7444 S 63.1 13.2 1035:20 java So that is 0.018t. Why not 18g, or something more accurate. I thought the credo was precision over readability. This is certainly not precise. Version-Release number of selected component (if applicable): How reproducible: easy Steps to Reproduce: 1. run a process that uses a lot of ram 2. run top 3. Actual results: 0.0nnt Expected results: nn.nG Additional info: Also in the man page at the bottom there is stated: 8. BUGS To report bugs, follow the instructions at: http://www.debian.org/Bugs/Reporting [^] Well that is funny, i have to go to the debian forums to report something in redhat/centos...
Hello Ellert, I have problems to reproduce this issue. These are outputs on RHEL-7.3 on x86_53 machine (first with default top output, second with scaling to human-readable form using 'e' key): default: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 12152 root 20 0 2574956 2.451g 0 S 0.0 87.5 0:00.69 memwaster with 'e' key: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 12152 root 20 0 2514.6m 2.155g 0.0m S 0.0 76.9 0:00.69 memwaster Did you use a custom config file for top (altering scale targets)? Also version of procps-ng package might help me trace this. rpm -qv procps-ng To the additional info you mentioned: The upstream version of procps (and procps-ng) package is created and maintained by Craig Small (and co-authored by Jim Warner) who also maintains these packages for Debian distribution. I believe Craig prefers this Debian forum as he checks it more frequently. Thank you and have a nice day.
I got it another round to find a reproducer. It seems like it only happens at the "Resident Memory" field and only if it exceeds 10GiB. Will be further investigated.
Got it. I found out it's another instance of "too little space in column" problem. The field for resident memory is set to be 6 characters long, which ##.###g does not fit, however 0.###t does. Fix is on its way.
Upstream NAKed the fix, because a new version containing the solution is said to be planned in future.
Associated: bz#1034466
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:0781