mirror of https://github.com/schoebel/mars
doc: add FAQ
This commit is contained in:
Thomas Schoebel-Theuer
parent
6b147f6707
commit
6c2b9c7c0f
|
|
@ -340,10 +340,44 @@ noprefix "false"
|
|||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
Finally, troubleshooting is explained in appendix
|
||||
Appendix
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand vref
|
||||
reference "chap:FAQ"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
||||
\end_inset
|
||||
|
||||
answers a few
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand nameref
|
||||
reference "chap:Handout-for-Midnight"
|
||||
reference "chap:FAQ"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
||||
\end_inset
|
||||
|
||||
s by the inventor of MARS.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
Appendix
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand vref
|
||||
reference "chap:Technical-Data-MARS"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
||||
\end_inset
|
||||
|
||||
contains some estimations for typical datacenter hardware, called
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand nameref
|
||||
reference "chap:Technical-Data-MARS"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
|
@ -31268,6 +31302,13 @@ Why
|
|||
systemd
|
||||
\family default
|
||||
?
|
||||
\begin_inset CommandInset label
|
||||
LatexCommand label
|
||||
name "subsec:Why-systemd?"
|
||||
|
||||
\end_inset
|
||||
|
||||
|
||||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
|
|
@ -31416,8 +31457,7 @@ Conflicts=
|
|||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
In the opinion of of author, systemd also has a few disadvantages, such
|
||||
as:
|
||||
In the opinion of the author, systemd has a few disadvantages, such as:
|
||||
\end_layout
|
||||
|
||||
\begin_layout Enumerate
|
||||
|
|
@ -35407,7 +35447,8 @@ df
|
|||
\family typewriter
|
||||
rest-space
|
||||
\family default
|
||||
but independently from it, this is also safeguarded and rounded down.
|
||||
but independently from it, this is also safeguarded and rounded down in
|
||||
units of GiB.
|
||||
Computational details may change in future MARS releases.
|
||||
\end_layout
|
||||
|
||||
|
|
@ -39656,6 +39697,392 @@ Test your scripts in failure scenarios!
|
|||
|
||||
\begin_layout Chapter
|
||||
\start_of_appendix
|
||||
FAQ
|
||||
\begin_inset CommandInset label
|
||||
LatexCommand label
|
||||
name "chap:FAQ"
|
||||
|
||||
\end_inset
|
||||
|
||||
|
||||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
Following answers are from Thomas Schöbel-Theuer, the inventor of the MARS
|
||||
components and the responsible maintainer of the out-of-tree OpenSource
|
||||
project under GPL.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
Q: Why are the units of
|
||||
\family typewriter
|
||||
marsdm view-rest-space
|
||||
\family default
|
||||
or
|
||||
\family typewriter
|
||||
marsdm view-total-space
|
||||
\family default
|
||||
in GiB?
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
A: This is for your protection.
|
||||
Contemporary servers have RAM up to TiB, not only GiB.
|
||||
In worst case, e.g.
|
||||
certain types of userspace bugs / attacks / intrusions, large parts of
|
||||
this RAM can be filled via the
|
||||
\emph on
|
||||
page cache of the kernel
|
||||
\emph default
|
||||
by (masses of) wild-running user processes
|
||||
\emph on
|
||||
like hell
|
||||
\emph default
|
||||
, e.g.
|
||||
\begin_inset space ~
|
||||
\end_inset
|
||||
|
||||
GiB/ms.
|
||||
Exceptional RAM usage / rates can trigger the
|
||||
\series bold
|
||||
OOM killer
|
||||
\series default
|
||||
, which in turn may
|
||||
\series bold
|
||||
\emph on
|
||||
randomly
|
||||
\emph default
|
||||
kill
|
||||
\series default
|
||||
almost anything,
|
||||
\series bold
|
||||
including daemons
|
||||
\series default
|
||||
|
||||
\begin_inset Foot
|
||||
status open
|
||||
|
||||
\begin_layout Plain Layout
|
||||
Exceptional fill rates are dangerous by itself.
|
||||
Neither the kernel nor MARS can
|
||||
\emph on
|
||||
reliably
|
||||
\emph default
|
||||
prevent such
|
||||
\begin_inset Quotes eld
|
||||
\end_inset
|
||||
|
||||
use cases
|
||||
\begin_inset Quotes erd
|
||||
\end_inset
|
||||
|
||||
.
|
||||
We altogether can only
|
||||
\emph on
|
||||
try
|
||||
\emph default
|
||||
to
|
||||
\emph on
|
||||
deal with it
|
||||
\emph default
|
||||
and to protect us
|
||||
\emph on
|
||||
as best as possible
|
||||
\emph default
|
||||
under certain conditions.
|
||||
Detailed high-frequency monitoring would be difficult in practice.
|
||||
Via units of GiB, I try to
|
||||
\emph on
|
||||
hint
|
||||
\emph default
|
||||
you on this large problem field.
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
Q: Why isn't
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
a daemon, but follows the process paradigm from UNIX?
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
A: Same story: reasonable protection against whoever kills processes for
|
||||
whatever reasons.
|
||||
Daemons can
|
||||
\series bold
|
||||
loose their runtime state
|
||||
\series default
|
||||
, and this may happen even in
|
||||
\emph on
|
||||
critical sections
|
||||
\emph default
|
||||
.
|
||||
Thus your
|
||||
\series bold
|
||||
(distributed) system(s) may get (widely) stuck
|
||||
\series default
|
||||
.
|
||||
More explanations are in section
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand nameref
|
||||
reference "subsec:Why-systemd?"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
||||
\end_inset
|
||||
|
||||
|
||||
\begin_inset CommandInset ref
|
||||
LatexCommand vpageref
|
||||
reference "subsec:Why-systemd?"
|
||||
plural "false"
|
||||
caps "false"
|
||||
noprefix "false"
|
||||
|
||||
\end_inset
|
||||
|
||||
.
|
||||
This can become
|
||||
\emph on
|
||||
evil
|
||||
\emph default
|
||||
when it happens just during an operational incident, requiring a (mass)
|
||||
failover, and just when you
|
||||
\emph on
|
||||
desperately
|
||||
\emph default
|
||||
want to
|
||||
\emph on
|
||||
rely
|
||||
\emph default
|
||||
on MARS (see also the CAP theorem).
|
||||
In contrast,
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
(as a component) is
|
||||
\series bold
|
||||
stateless
|
||||
\begin_inset Foot
|
||||
status open
|
||||
|
||||
\begin_layout Plain Layout
|
||||
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
does not
|
||||
\emph on
|
||||
functionally
|
||||
\emph default
|
||||
depend on network traffic, at least for non-network operations.
|
||||
The
|
||||
\emph on
|
||||
kernel(!) and its GPLed code
|
||||
\emph default
|
||||
will
|
||||
\emph on
|
||||
automatically
|
||||
\emph default
|
||||
cleanup most remains upon process kill.
|
||||
|
||||
\series bold
|
||||
Repeatability
|
||||
\series default
|
||||
: afterwards, the next
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
invocation should not depend on former (OOM-like) problems.
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
|
||||
\series default
|
||||
|
||||
\emph on
|
||||
between
|
||||
\emph default
|
||||
invocations.
|
||||
Thus your system(s) can
|
||||
\series bold
|
||||
recover more easily
|
||||
\series default
|
||||
.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
Q: Can I create a fork of MARS?
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
A: Yes, given that you
|
||||
\series bold
|
||||
conform to the license
|
||||
\series default
|
||||
.
|
||||
Please read the GPL, and
|
||||
\series bold
|
||||
obey the GPL
|
||||
\series default
|
||||
.
|
||||
This is not only about kernel code.
|
||||
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
will remain under GPL.
|
||||
I have
|
||||
\emph on
|
||||
created
|
||||
\emph default
|
||||
it under European personal rights, and even during my precious spare time.
|
||||
Personally, I will
|
||||
\emph on
|
||||
not
|
||||
\emph default
|
||||
publish anything which is
|
||||
\emph on
|
||||
not
|
||||
\emph default
|
||||
under an OSI and/or OIN-compliant license.
|
||||
\begin_inset Newline newline
|
||||
\end_inset
|
||||
|
||||
In addition, it is
|
||||
\emph on
|
||||
more than wise
|
||||
\emph default
|
||||
to obey the customs of the OpenSource
|
||||
\series bold
|
||||
community
|
||||
\series default
|
||||
.
|
||||
I
|
||||
\series bold
|
||||
firmly recommend
|
||||
\series default
|
||||
to
|
||||
\emph on
|
||||
fully
|
||||
\emph default
|
||||
read and to
|
||||
\emph on
|
||||
fully
|
||||
\emph default
|
||||
understand and to obey
|
||||
\family sans
|
||||
\series bold
|
||||
Homesteading the Noosphere
|
||||
\family default
|
||||
\series default
|
||||
(a ground-breaking essay from Eric Raymond, see original web document at
|
||||
|
||||
\begin_inset Flex URL
|
||||
status open
|
||||
|
||||
\begin_layout Plain Layout
|
||||
|
||||
http://catb.org/~esr/writings/homesteading/homesteading/
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
).
|
||||
It may take several days or weeks to
|
||||
\emph on
|
||||
really
|
||||
\emph default
|
||||
understand this, and its consequences.
|
||||
It is
|
||||
\emph on
|
||||
more than unwise
|
||||
\emph default
|
||||
to break important basics from there.
|
||||
In addition, it is mandatory to obey the written and non-written rules
|
||||
from the Linux kernel community.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
Q: Will MARS become part of the official Linux kernel as published at
|
||||
\begin_inset Flex URL
|
||||
status open
|
||||
|
||||
\begin_layout Plain Layout
|
||||
|
||||
www.kernel.org
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
?
|
||||
\end_layout
|
||||
|
||||
\begin_layout Labeling
|
||||
\labelwidthstring 00.00.0000
|
||||
A: Hopefully.
|
||||
The decision is up to Linus Torvalds and the kernel community.
|
||||
If possible, I will try to do my best, in order to get it there
|
||||
\begin_inset Foot
|
||||
status open
|
||||
|
||||
\begin_layout Plain Layout
|
||||
Hopefully, the
|
||||
\emph on
|
||||
out-of-tree
|
||||
\emph default
|
||||
kernel part of MARS will no longer be needed by anyone in the long term.
|
||||
Any (professional) support for the MARS kernel module, and also for
|
||||
\family typewriter
|
||||
marsadm
|
||||
\family default
|
||||
, should hopefully migrate to the official Linux kernel and/or to the OpenSource
|
||||
community of all interested Linux distros, in a reasonable timeframe.
|
||||
\end_layout
|
||||
|
||||
\end_inset
|
||||
|
||||
.
|
||||
After acceptance, I am willing to support it further as a member of the
|
||||
community, up to my death, under my personal constraints after my retirement.
|
||||
I am also planning to continuously support the community, and to consent
|
||||
to passing over the baton to somebody
|
||||
\emph on
|
||||
from the Linux kernel community
|
||||
\emph default
|
||||
.
|
||||
My current problem is lack of time.
|
||||
I estimate that I will need more than 70% of my precious working hours
|
||||
for getting it into the kernel.
|
||||
Please help me.
|
||||
Please talk
|
||||
\emph on
|
||||
directly
|
||||
\emph default
|
||||
with me about perspectives.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Standard
|
||||
This FAQ will be continued / updated only when necessary, and according
|
||||
to my time constraints.
|
||||
\end_layout
|
||||
|
||||
\begin_layout Chapter
|
||||
Technical Data MARS
|
||||
\begin_inset CommandInset label
|
||||
LatexCommand label
|
||||
|
|
|
|||
Loading…
Reference in New Issue