I have a pretty nice workstation, with 8 cores and 32G RAM.
A stupid little python program just killed it by consuming all the memory, driving it into swapping hell.
I couldn't even ssh into it - I started up ssh and went for coffee. On return I still didn't have a prompt. Had to power cycle it.
This is F23, everything is setup default regarding kernel memory policies, etc.
So, can we configure things to give a better experience? Can we make this default?
On 1 December 2015 at 15:19, Neal Becker ndbecker2@gmail.com wrote:
I have a pretty nice workstation, with 8 cores and 32G RAM.
A stupid little python program just killed it by consuming all the memory, driving it into swapping hell.
I couldn't even ssh into it - I started up ssh and went for coffee. On return I still didn't have a prompt. Had to power cycle it.
This is F23, everything is setup default regarding kernel memory policies, etc.
So, can we configure things to give a better experience? Can we make this default?
Maybe one for the devel list where they can do something about it. I haven't actually checked what the current Fedora policies are, since at work I use RHEL. I'd have thought the oom killer would get this. Really taking out all the memory with no swap available seems more likely to kill a system, possibly memory use can expand too fast, but on systems with some swap I've rarely seen things get to the point you can't get a virtual terminal up.
Ian Malone wrote:
On 1 December 2015 at 15:19, Neal Becker ndbecker2@gmail.com wrote:
I have a pretty nice workstation, with 8 cores and 32G RAM.
A stupid little python program just killed it by consuming all the memory, driving it into swapping hell.
I couldn't even ssh into it - I started up ssh and went for coffee. On return I still didn't have a prompt. Had to power cycle it.
This is F23, everything is setup default regarding kernel memory policies, etc.
So, can we configure things to give a better experience? Can we make this default?
Maybe one for the devel list where they can do something about it. I haven't actually checked what the current Fedora policies are, since at work I use RHEL. I'd have thought the oom killer would get this. Really taking out all the memory with no swap available seems more likely to kill a system, possibly memory use can expand too fast, but on systems with some swap I've rarely seen things get to the point you can't get a virtual terminal up. cat /proc/meminfo
MemTotal: 32902244 kB MemFree: 24712892 kB MemAvailable: 28580148 kB Buffers: 2996 kB Cached: 3769828 kB SwapCached: 0 kB Active: 5229044 kB Inactive: 2420832 kB Active(anon): 3878352 kB Inactive(anon): 36628 kB Active(file): 1350692 kB Inactive(file): 2384204 kB Unevictable: 16 kB Mlocked: 16 kB SwapTotal: 16400376 kB SwapFree: 16400376 kB Dirty: 0 kB Writeback: 0 kB AnonPages: 3877468 kB Mapped: 307056 kB Shmem: 37928 kB Slab: 295500 kB SReclaimable: 218208 kB SUnreclaim: 77292 kB KernelStack: 9264 kB PageTables: 41680 kB NFS_Unstable: 0 kB Bounce: 0 kB WritebackTmp: 0 kB CommitLimit: 32851496 kB Committed_AS: 6566200 kB VmallocTotal: 34359738367 kB VmallocUsed: 190544 kB VmallocChunk: 34359531516 kB HardwareCorrupted: 0 kB AnonHugePages: 0 kB HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 2048 kB DirectMap4k: 194712 kB DirectMap2M: 6053888 kB DirectMap1G: 28311552 kB
On 12/01/2015 07:19 AM, Neal Becker wrote:
I have a pretty nice workstation, with 8 cores and 32G RAM.
A stupid little python program just killed it by consuming all the memory, driving it into swapping hell.
I couldn't even ssh into it - I started up ssh and went for coffee. On return I still didn't have a prompt. Had to power cycle it.
This is F23, everything is setup default regarding kernel memory policies, etc.
So, can we configure things to give a better experience? Can we make this default?
You can sure edit /etc/security/limits.conf and bugger values there to limit it based on user IDs. ---------------------------------------------------------------------- - Rick Stevens, Systems Engineer, AllDigital ricks@alldigital.com - - AIM/Skype: therps2 ICQ: 226437340 Yahoo: origrps2 - - - - IGNORE that man behind the keyboard! - - - The Wizard of OS - ----------------------------------------------------------------------
On 1 December 2015 at 18:47, Neal Becker ndbecker2@gmail.com wrote:
Ian Malone wrote:
On 1 December 2015 at 15:19, Neal Becker ndbecker2@gmail.com wrote:
I have a pretty nice workstation, with 8 cores and 32G RAM.
A stupid little python program just killed it by consuming all the memory, driving it into swapping hell.
I couldn't even ssh into it - I started up ssh and went for coffee. On return I still didn't have a prompt. Had to power cycle it.
Maybe one for the devel list where they can do something about it. I haven't actually checked what the current Fedora policies are, since at work I use RHEL. I'd have thought the oom killer would get this. Really taking out all the memory with no swap available seems more likely to kill a system, possibly memory use can expand too fast, but on systems with some swap I've rarely seen things get to the point you can't get a virtual terminal up.
cat /proc/meminfo
<snip>
Well, first time for everything. Can you share what the program was? There are examples on stackoverflow which intentionally do this kind of thing, but doing it without meaning to is impressive. For my own interest, is your swap on SSD?
-
Ian Malone -
Neal Becker -
Rick Stevens