y**b
发帖数: 73 | 1 我们学校搞了一个超大超强的cluster用于计算需要，可惜管理得太差。几乎隔几天就
可以收到
email说系统不行了，过几个小时又说系统修复了。不过我还是时不时在上面run一些
job。
今天收到管理员的一封信，笑死我了。我猜我可以把这个写进我的resume，也算是成就
之一。
信是这么说的
Dear ***,
Today at 4:30pm you ran an interactive job. This job ran on heroint2.
For some reason the I/O pattern of your job killed the file server that
serves our home directory. As you can imagine we are *very* interested
in finding out what you were doing and to see if we can reproduce what
happened. We would not have expected a single job to be abl |
p*******o
发帖数: 2726 | 2 Niu!
【在 y**b 的大作中提到】
: 我们学校搞了一个超大超强的cluster用于计算需要，可惜管理得太差。几乎隔几天就
: 可以收到
: email说系统不行了，过几个小时又说系统修复了。不过我还是时不时在上面run一些
: job。
: 今天收到管理员的一封信，笑死我了。我猜我可以把这个写进我的resume，也算是成就
: 之一。
: 信是这么说的
: Dear ***,
: Today at 4:30pm you ran an interactive job. This job ran on heroint2.
: For some reason the I/O pattern of your job killed the file server that
|
A*****n
发帖数: 243 | 3 job过于频繁地从工作结点访问front end node？
【在 y**b 的大作中提到】
: 我们学校搞了一个超大超强的cluster用于计算需要，可惜管理得太差。几乎隔几天就
: 可以收到
: email说系统不行了，过几个小时又说系统修复了。不过我还是时不时在上面run一些
: job。
: 今天收到管理员的一封信，笑死我了。我猜我可以把这个写进我的resume，也算是成就
: 之一。
: 信是这么说的
: Dear ***,
: Today at 4:30pm you ran an interactive job. This job ran on heroint2.
: For some reason the I/O pattern of your job killed the file server that
|
s*r
发帖数: 2757 | 4 你是不是在主node上运行大程序呀
大量的i/o其实不是一件好事
【在 y**b 的大作中提到】
: 我们学校搞了一个超大超强的cluster用于计算需要，可惜管理得太差。几乎隔几天就
: 可以收到
: email说系统不行了，过几个小时又说系统修复了。不过我还是时不时在上面run一些
: job。
: 今天收到管理员的一封信，笑死我了。我猜我可以把这个写进我的resume，也算是成就
: 之一。
: 信是这么说的
: Dear ***,
: Today at 4:30pm you ran an interactive job. This job ran on heroint2.
: For some reason the I/O pattern of your job killed the file server that
|
y**b
发帖数: 73 | 5 My code is super simple, and it's run according to the submission rule of
the cluster. Any heavy job in the main node will be killed right away. I do
not have extensive I/O at all. |
l*****k
发帖数: 587 | 6 I find R does not release memory after it uses it, I recently run it
on a winxp with 2G memory, which takes up memory while running, but the
PC becomes super slow even after the job is finished, I have to power it
down and reboot, of course I lost all result I ever generated.
I think it does the same on linux, anyone have a fix?
【在 y**b 的大作中提到】
: 我们学校搞了一个超大超强的cluster用于计算需要，可惜管理得太差。几乎隔几天就
: 可以收到
: email说系统不行了，过几个小时又说系统修复了。不过我还是时不时在上面run一些
: job。
: 今天收到管理员的一封信，笑死我了。我猜我可以把这个写进我的resume，也算是成就
: 之一。
: 信是这么说的
: Dear ***,
: Today at 4:30pm you ran an interactive job. This job ran on heroint2.
: For some reason the I/O pattern of your job killed the file server that
|
s*r
发帖数: 2757 | 7 i guess that was because R uses a lot of virtual memory
on hard drive. after exiting R, the system still used those
virtual memory, which slowed everything
【在 l*****k 的大作中提到】
: I find R does not release memory after it uses it, I recently run it
: on a winxp with 2G memory, which takes up memory while running, but the
: PC becomes super slow even after the job is finished, I have to power it
: down and reboot, of course I lost all result I ever generated.
: I think it does the same on linux, anyone have a fix?
|
s*r
发帖数: 2757 | 8 ok. that is interesting then
do
【在 y**b 的大作中提到】
: My code is super simple, and it's run according to the submission rule of
: the cluster. Any heavy job in the main node will be killed right away. I do
: not have extensive I/O at all.
|
h******a
发帖数: 198 | 9 我也觉得R 好像不会内存整理。自己用个软件整理一下吧
【在 l*****k 的大作中提到】
: I find R does not release memory after it uses it, I recently run it
: on a winxp with 2G memory, which takes up memory while running, but the
: PC becomes super slow even after the job is finished, I have to power it
: down and reboot, of course I lost all result I ever generated.
: I think it does the same on linux, anyone have a fix?
|
g********r
发帖数: 8017 | 10 gc()
【在 l*****k 的大作中提到】
: I find R does not release memory after it uses it, I recently run it
: on a winxp with 2G memory, which takes up memory while running, but the
: PC becomes super slow even after the job is finished, I have to power it
: down and reboot, of course I lost all result I ever generated.
: I think it does the same on linux, anyone have a fix?
|
q**j
发帖数: 10612 | 11 there is a function to release the memory. forget the name. you can find it
in the little book i wrote though.
【在 l*****k 的大作中提到】
: I find R does not release memory after it uses it, I recently run it
: on a winxp with 2G memory, which takes up memory while running, but the
: PC becomes super slow even after the job is finished, I have to power it
: down and reboot, of course I lost all result I ever generated.
: I think it does the same on linux, anyone have a fix?
|
g********r
发帖数: 8017 | 12 garbage collection. gc(). I had to use it many times.
it
【在 q**j 的大作中提到】
: there is a function to release the memory. forget the name. you can find it
: in the little book i wrote though.
|
