Proposal: Core dump in docker #19289

zhangjianfnst · 2016-01-13T09:02:17Z

In the current docker, we can only use the system global config of /proc/sys/kernel/core_pattern to specify the output place of core dump in container. There are several problems in this way:

It maybe affect the original core dump behaviour config for host.
It can't specify the output place of core dump independently for every container.
The specified output place of core dump is in the container, but there are no tools (gdb e.g.) to investigate the core dump file in the container, and the core dump file may be too large. So it's better
to store the core dump out of the container.

One proposal of implementation steps:

Introduce "--coredump-helper" option in docker daemon, this option enables and specifies core dump helper program.
E.g.:
docker daemon --coredump-helper=/usr/libexec/docker-coredump-helper

docker-coredump-helper is an executable program built from a project of docker-coredump-helper (it doesn't exsist now, also need to be implemented).

Process steps:
1. Require root
2. Check /usr/libexec/docker-coredump-helper
3. Copy /proc/sys/kernel/core_pattern to /var/run/docker/core_patterns/host
4. Set /proc/sys/kernel/core_pattern with "|/usr/libexec/docker-coredump-helper %c %d %e %E %g %h %i %I %p %P %s %t %u"
If the value of "--coredump-helper" isn't specified, docker daemon need not do the above steps.
Introduce "--core-pattern" option in docker run/create, this option specifies the container core pattern.
E.g.:
a) docker run --core-pattern="#/var/core.%s.%p.%e" [IMAGE] [CMD] // path with '#' means it would be in host mnt namespace
b) docker run --core-pattern="/var/core.%s.%p.%e" [IMAGE] [CMD] // path is in the container mnt namespace
c) docker run [IMAGE] [CMD] // if "--core-pattern" not specified, default as: "#/var/run/docker/core_files/[ContainerID]/core.%p"

Process steps:
1. Check "--core-pattern" option
2. Save core-pattern value to /var/run/docker/core_patterns/[ContainerID]
```
Tree of core_patterns:
 /var/run/docker/core_patterns/
 ├── host
 ├── [ContainerID1]
 ├── [ContainerID2]
 ├── ...
 └── [ContainerIDn]
```
1. When core dump is triggered, the docker-coredump-helper is executed. This program can judge the core dump source process and dispatch the core dump file according the config of /var/run/docker/core_patterns/*.

The text was updated successfully, but these errors were encountered:

zhangjianfnst · 2016-01-15T01:28:41Z

@cpuguy83 @moxiegirl @thaJeztah @vdemeester Hi, what do you think about this? Could you please give some comments?

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

LK4D4 · 2016-09-14T18:17:14Z

This looks like nice thing to have.

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

sanmai-NL · 2016-10-23T21:08:39Z

Supporting, since it appears impossible to use a host data volume as a custom core_pattern file instead.

Docker Engine 1.12.2:

oci runtime error: rootfs_linux.go:53: mounting "/proc/sys/kernel/core_pattern" to rootfs 
"/var/lib/docker/devicemapper/mnt/28be96a0d47790bbe9560482a100bc475326f0f2cfb3ba6393ddf0d295b0eaf4/rootfs" caused 
"\\"/var/lib/docker/devicemapper/mnt/28be96a0d47790bbe9560482a100bc475326f0f2cfb3ba6393ddf0d295b0eaf4/rootfs/proc/sys/kernel/core_pattern\\" 
cannot be mounted because it is located inside \\"/proc\\""'

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

playground · 2017-12-14T19:03:33Z

What is the best way to get the core dump files onto my local machine from remote container so I can examine the files? The files are huge.

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

kajmagnus · 2018-08-06T03:55:12Z

@playground What about installing gdb on the remote machine? (the host or the container, both could work I would think) and running gdb ... backtrace remotely? Or maybe mount the /tmp/core-dump-directory/ in an object store like Amazon S3 or Google Cloud's object storage.

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

WanLinghao · 2019-04-24T06:29:02Z

Hi, everyone, do we have any progress about this issue?

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

adelca · 2020-03-16T23:06:58Z

2020 - it is still a struggle to get a core dump without having to start container in privilege mode and impacting host config and other PODs... coronavirus will get us all before we can analyse those juicy core dumps waiting for us

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com> +++ ve, coredump: Fix usercopy warning. Usercopy WARNs about copying ve.core_pattern to user space: vzctl exec e2783e5e-840e-4c9a-9ef2-7ead73afb81c cat /proc/sys/kernel/core_pattern Call Trace: __check_object_size+0x134/0x160 proc_dostring+0x164/0x200 proc_dostring_coredump_virtual+0xb2/0xd0 proc_dostring+0x200/0x200 proc_sys_call_handler+0xa7/0xf0 vfs_read+0x9d/0x150 ksys_read+0x4f/0xb0 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x65/0xca Whitelist core_pattern in ve_struct kmem_cache since copying core_pattern is allowed by design. Fixes: e931118 ("ve: Add ve cgroup and ve_hook subsys") https://jira.sw.ru/browse/PSBM-106216 Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com> vz9 change: merge with usercopy warning fixup for ve->core_pattern (cherry picked from vz8 commit f8c44aa) Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>

kuvaldini · 2021-12-10T16:06:47Z

Too much effort. And therefore not implemented during 5yrs. For me it would be just enough

docker run --core-pattern="WORKS exactly kernel.core_pattern=" [IMAGE] [CMD]

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com> +++ ve, coredump: Fix usercopy warning. Usercopy WARNs about copying ve.core_pattern to user space: vzctl exec e2783e5e-840e-4c9a-9ef2-7ead73afb81c cat /proc/sys/kernel/core_pattern Call Trace: __check_object_size+0x134/0x160 proc_dostring+0x164/0x200 proc_dostring_coredump_virtual+0xb2/0xd0 proc_dostring+0x200/0x200 proc_sys_call_handler+0xa7/0xf0 vfs_read+0x9d/0x150 ksys_read+0x4f/0xb0 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x65/0xca Whitelist core_pattern in ve_struct kmem_cache since copying core_pattern is allowed by design. Fixes: e931118 ("ve: Add ve cgroup and ve_hook subsys") https://jira.sw.ru/browse/PSBM-106216 Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com> vz9 change: merge with usercopy warning fixup for ve->core_pattern (cherry picked from vz8 commit f8c44aa) Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>

1) put core_pattern on ve_struct and add virtualized proc_handler using macro sysctl_virtual() 2) use in-container khelper kthread for collecting cores We need support of pipe'ing coredump in CT. If process crashes in CT and kernel.core_pattern begins with "|" we get error: CT: CT-101-bash-4.2# sysctl kernel.core_pattern kernel.core_pattern = |/root/core_test %p UID=%u GID=%g sig=%s CT-101-bash-4.2# sleep 100 & [1] 445 CT-101-bash-4.2# kill -11 445 CT-101-bash-4.2# [1]+ Segmentation fault sleep 100 CT-101-bash-4.2# ll /root total 8 rwxr-xr-x 1 root root 7424 Jan 27 17:28 core_test Host: [root@s143 ~]# less /var/log/messages Jan 28 10:40:42 s143 kernel: [87698.969582] Core dump to |/root/core_test 445 UID=0 GID=0 sig=11 pipe failed Also docker is going to use pipe'ing in coredumps functionality: moby/moby#19289 https://jira.sw.ru/browse/PSBM-43596 v2: split sysctl_virtual in separate patch Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com> Signed-off-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com> +++ ve, coredump: Fix usercopy warning. Usercopy WARNs about copying ve.core_pattern to user space: vzctl exec e2783e5e-840e-4c9a-9ef2-7ead73afb81c cat /proc/sys/kernel/core_pattern Call Trace: __check_object_size+0x134/0x160 proc_dostring+0x164/0x200 proc_dostring_coredump_virtual+0xb2/0xd0 proc_dostring+0x200/0x200 proc_sys_call_handler+0xa7/0xf0 vfs_read+0x9d/0x150 ksys_read+0x4f/0xb0 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x65/0xca Whitelist core_pattern in ve_struct kmem_cache since copying core_pattern is allowed by design. Fixes: e931118 ("ve: Add ve cgroup and ve_hook subsys") https://jira.sw.ru/browse/PSBM-106216 Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com> vz9 change: merge with usercopy warning fixup for ve->core_pattern (cherry picked from vz8 commit f8c44aa) Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Feature: sysctl: virtualize "kernel.core_pattern"

zhangjianfnst mentioned this issue Jan 15, 2016

Documentation: Make clear instructions for getting a core file, when container crashes #11740

Open

thaJeztah added the kind/feature Functionality or other elements that the project doesn't currently have. Features are new and shiny label Jan 15, 2016

LK4D4 added the area/runtime label Sep 14, 2016

GuillaumeDerval mentioned this issue Sep 22, 2016

cpp: Add gdb INGInious/containers#11

Merged

AnthonyMastrean mentioned this issue Jul 10, 2019

Core dumps fill the host's disk dotnet/dotnet-docker#1215

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Proposal: Core dump in docker #19289

Proposal: Core dump in docker #19289

zhangjianfnst commented Jan 13, 2016

zhangjianfnst commented Jan 15, 2016

LK4D4 commented Sep 14, 2016

sanmai-NL commented Oct 23, 2016

playground commented Dec 14, 2017

kajmagnus commented Aug 6, 2018 •

edited

Loading

WanLinghao commented Apr 24, 2019

adelca commented Mar 16, 2020

kuvaldini commented Dec 10, 2021

Proposal: Core dump in docker #19289

Proposal: Core dump in docker #19289

Comments

zhangjianfnst commented Jan 13, 2016

zhangjianfnst commented Jan 15, 2016

LK4D4 commented Sep 14, 2016

sanmai-NL commented Oct 23, 2016

playground commented Dec 14, 2017

kajmagnus commented Aug 6, 2018 • edited Loading

WanLinghao commented Apr 24, 2019

adelca commented Mar 16, 2020

kuvaldini commented Dec 10, 2021

kajmagnus commented Aug 6, 2018 •

edited

Loading