Hide Forgot
Description of problem: nautilus (and caribou) consuming too much memory when man block devices are present. I was testing creating many LVs on a system with running gnome session. I expected memory to be stressed a bit but did not expect nautilus (and caribou) consuming 6.46GB (and 2.86GB) or RAM. This might be a problem on machines running for example docker, where density of running 1000+ containers is the goal. As some of administration tools are using GUI it is not impossible and could affect customers. Version-Release number of selected component (if applicable): nautilus-3.14.3-7.el7.x86_64 caribou-0.4.16-1.el7.x86_64 How reproducible: ? Steps to Reproduce: 1. for i in $(seq 1500); do lvcreate -s $VG/$THINLV; done Actual results: Wasted 9+ GB or RAM. Expected results: There should be some saner limit what file manager consumes. Especially when doing "nothing". Additional info: > top - 13:22:11 up 26 days, 2:29, 8 users, load average: 0.18, 0.29, 0.37 > Tasks: 3776 total, 1 running, 3770 sleeping, 5 stopped, 0 zombie > %Cpu(s): 27.9 us, 4.3 sy, 0.2 ni, 67.6 id, 0.1 wa, 0.0 hi, 0.0 si, 0.0 st > KiB Mem : 15889328 total, 165116 free, 14954932 used, 769280 buff/cache > KiB Swap: 8388604 total, 1828152 free, 6560452 used. 204764 avail Mem > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 19891 mrtianc 20 0 8001360 6.462g 1248 S 0.0 42.6 24:14.70 nautilus > 19721 mrtianc 20 0 3839744 2.866g 904 S 0.0 18.9 14:04.27 caribou > 25805 mcsontos 20 0 2028732 658272 15216 S 0.0 4.1 9:00.27 firefox
Adding information from caribou Bug 1331779: Even deactivating LVs causes both nautilus and caribou leaking a lot. After restarting gnome, the programs were fine, until I started manipulating the LVs. After deactivating 1000LVs it now looks like this: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 28616 mrtianc 20 0 2616316 1.698g 16968 S 0.0 11.2 5:59.90 nautilus 27886 mrtianc 20 0 1314972 751520 5044 S 0.0 4.7 2:58.94 caribou Nautilus and caribou were running at full speed for 6 minutes, nautilus consuming 100% CPU, and caribou "only" 50%. Reactivating all these 1000 LVs gets both applications running again. The fact deactivating devices is causing memory going up suggests there is a huge leak - over 1MB in nautilus and 0.7MB in caribou for each activated/deactivated LV. Activating the volumes back the leaks are much worse: 4M in nautilus and 1.8M in caribou per LV. > 28616 mrtianc 20 0 6490804 5.334g 3052 S 0.0 35.2 20:20.58 nautilus > 27886 mrtianc 20 0 3109652 2.407g 908 S 0.0 15.9 10:06.78 caribou
I don't have a suitable environment to test this right now, can you please confirm that this is still valid for RHEL 7.7?
Finally, I was able to run the reproducer on RHEL 7.7, but I don't see any memory leaks with nautilus-3.26.3.1-6.el7.x86_64, thus closing.