Commit e4f08c46 authored by Iustin Pop's avatar Iustin Pop
Browse files

Initial import

This is the initial import of release 0.0.3.
parents
GNU GENERAL PUBLIC LICENSE
Version 2, June 1991
Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The licenses for most software are designed to take away your
freedom to share and change it. By contrast, the GNU General Public
License is intended to guarantee your freedom to share and change free
software--to make sure the software is free for all its users. This
General Public License applies to most of the Free Software
Foundation's software and to any other program whose authors commit to
using it. (Some other Free Software Foundation software is covered by
the GNU Lesser General Public License instead.) You can apply it to
your programs, too.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
this service if you wish), that you receive source code or can get it
if you want it, that you can change the software or use pieces of it
in new free programs; and that you know you can do these things.
To protect your rights, we need to make restrictions that forbid
anyone to deny you these rights or to ask you to surrender the rights.
These restrictions translate to certain responsibilities for you if you
distribute copies of the software, or if you modify it.
For example, if you distribute copies of such a program, whether
gratis or for a fee, you must give the recipients all the rights that
you have. You must make sure that they, too, receive or can get the
source code. And you must show them these terms so they know their
rights.
We protect your rights with two steps: (1) copyright the software, and
(2) offer you this license which gives you legal permission to copy,
distribute and/or modify the software.
Also, for each author's protection and ours, we want to make certain
that everyone understands that there is no warranty for this free
software. If the software is modified by someone else and passed on, we
want its recipients to know that what they have is not the original, so
that any problems introduced by others will not reflect on the original
authors' reputations.
Finally, any free program is threatened constantly by software
patents. We wish to avoid the danger that redistributors of a free
program will individually obtain patent licenses, in effect making the
program proprietary. To prevent this, we have made it clear that any
patent must be licensed for everyone's free use or not licensed at all.
The precise terms and conditions for copying, distribution and
modification follow.
GNU GENERAL PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
0. This License applies to any program or other work which contains
a notice placed by the copyright holder saying it may be distributed
under the terms of this General Public License. The "Program", below,
refers to any such program or work, and a "work based on the Program"
means either the Program or any derivative work under copyright law:
that is to say, a work containing the Program or a portion of it,
either verbatim or with modifications and/or translated into another
language. (Hereinafter, translation is included without limitation in
the term "modification".) Each licensee is addressed as "you".
Activities other than copying, distribution and modification are not
covered by this License; they are outside its scope. The act of
running the Program is not restricted, and the output from the Program
is covered only if its contents constitute a work based on the
Program (independent of having been made by running the Program).
Whether that is true depends on what the Program does.
1. You may copy and distribute verbatim copies of the Program's
source code as you receive it, in any medium, provided that you
conspicuously and appropriately publish on each copy an appropriate
copyright notice and disclaimer of warranty; keep intact all the
notices that refer to this License and to the absence of any warranty;
and give any other recipients of the Program a copy of this License
along with the Program.
You may charge a fee for the physical act of transferring a copy, and
you may at your option offer warranty protection in exchange for a fee.
2. You may modify your copy or copies of the Program or any portion
of it, thus forming a work based on the Program, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:
a) You must cause the modified files to carry prominent notices
stating that you changed the files and the date of any change.
b) You must cause any work that you distribute or publish, that in
whole or in part contains or is derived from the Program or any
part thereof, to be licensed as a whole at no charge to all third
parties under the terms of this License.
c) If the modified program normally reads commands interactively
when run, you must cause it, when started running for such
interactive use in the most ordinary way, to print or display an
announcement including an appropriate copyright notice and a
notice that there is no warranty (or else, saying that you provide
a warranty) and that users may redistribute the program under
these conditions, and telling the user how to view a copy of this
License. (Exception: if the Program itself is interactive but
does not normally print such an announcement, your work based on
the Program is not required to print an announcement.)
These requirements apply to the modified work as a whole. If
identifiable sections of that work are not derived from the Program,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works. But when you
distribute the same sections as part of a whole which is a work based
on the Program, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote it.
Thus, it is not the intent of this section to claim rights or contest
your rights to work written entirely by you; rather, the intent is to
exercise the right to control the distribution of derivative or
collective works based on the Program.
In addition, mere aggregation of another work not based on the Program
with the Program (or with a work based on the Program) on a volume of
a storage or distribution medium does not bring the other work under
the scope of this License.
3. You may copy and distribute the Program (or a work based on it,
under Section 2) in object code or executable form under the terms of
Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable
source code, which must be distributed under the terms of Sections
1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three
years, to give any third party, for a charge no more than your
cost of physically performing source distribution, a complete
machine-readable copy of the corresponding source code, to be
distributed under the terms of Sections 1 and 2 above on a medium
customarily used for software interchange; or,
c) Accompany it with the information you received as to the offer
to distribute corresponding source code. (This alternative is
allowed only for noncommercial distribution and only if you
received the program in object code or executable form with such
an offer, in accord with Subsection b above.)
The source code for a work means the preferred form of the work for
making modifications to it. For an executable work, complete source
code means all the source code for all modules it contains, plus any
associated interface definition files, plus the scripts used to
control compilation and installation of the executable. However, as a
special exception, the source code distributed need not include
anything that is normally distributed (in either source or binary
form) with the major components (compiler, kernel, and so on) of the
operating system on which the executable runs, unless that component
itself accompanies the executable.
If distribution of executable or object code is made by offering
access to copy from a designated place, then offering equivalent
access to copy the source code from the same place counts as
distribution of the source code, even though third parties are not
compelled to copy the source along with the object code.
4. You may not copy, modify, sublicense, or distribute the Program
except as expressly provided under this License. Any attempt
otherwise to copy, modify, sublicense or distribute the Program is
void, and will automatically terminate your rights under this License.
However, parties who have received copies, or rights, from you under
this License will not have their licenses terminated so long as such
parties remain in full compliance.
5. You are not required to accept this License, since you have not
signed it. However, nothing else grants you permission to modify or
distribute the Program or its derivative works. These actions are
prohibited by law if you do not accept this License. Therefore, by
modifying or distributing the Program (or any work based on the
Program), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Program or works based on it.
6. Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties to
this License.
7. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Program at all. For example, if a patent
license would not permit royalty-free redistribution of the Program by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Program.
If any portion of this section is held invalid or unenforceable under
any particular circumstance, the balance of the section is intended to
apply and the section as a whole is intended to apply in other
circumstances.
It is not the purpose of this section to induce you to infringe any
patents or other property right claims or to contest validity of any
such claims; this section has the sole purpose of protecting the
integrity of the free software distribution system, which is
implemented by public license practices. Many people have made
generous contributions to the wide range of software distributed
through that system in reliance on consistent application of that
system; it is up to the author/donor to decide if he or she is willing
to distribute software through any other system and a licensee cannot
impose that choice.
This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.
8. If the distribution and/or use of the Program is restricted in
certain countries either by patents or by copyrighted interfaces, the
original copyright holder who places the Program under this License
may add an explicit geographical distribution limitation excluding
those countries, so that distribution is permitted only in or among
countries not thus excluded. In such case, this License incorporates
the limitation as if written in the body of this License.
9. The Free Software Foundation may publish revised and/or new versions
of the General Public License from time to time. Such new versions will
be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the Program
specifies a version number of this License which applies to it and "any
later version", you have the option of following the terms and conditions
either of that version or of any later version published by the Free
Software Foundation. If the Program does not specify a version number of
this License, you may choose any version ever published by the Free Software
Foundation.
10. If you wish to incorporate parts of the Program into other free
programs whose distribution conditions are different, write to the author
to ask for permission. For software which is copyrighted by the Free
Software Foundation, write to the Free Software Foundation; we sometimes
make exceptions for this. Our decision will be guided by the two goals
of preserving the free status of all derivatives of our free software and
of promoting the sharing and reuse of software generally.
NO WARRANTY
11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
REPAIR OR CORRECTION.
12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
Also add information on how to contact you by electronic and paper mail.
If the program is interactive, make it output a short notice like this
when it starts in an interactive mode:
Gnomovision version 69, Copyright (C) year name of author
Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
This is free software, and you are welcome to redistribute it
under certain conditions; type `show c' for details.
The hypothetical commands `show w' and `show c' should show the appropriate
parts of the General Public License. Of course, the commands you use may
be called something other than `show w' and `show c'; they could even be
mouse-clicks or menu items--whatever suits your program.
You should also get your employer (if you work as a programmer) or your
school, if any, to sign a "copyright disclaimer" for the program, if
necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the program
`Gnomovision' (which makes passes at compilers) written by James Hacker.
<signature of Ty Coon>, 1 April 1989
Ty Coon, President of Vice
This General Public License does not permit incorporating your program into
proprietary programs. If your program is a subroutine library, you may
consider it more useful to permit linking proprietary applications with the
library. If this is what you want to do, use the GNU Lesser General
Public License instead of this License.
HSRCS := $(wildcard src/*.hs)
HDDIR = apidoc
# Haskell rules
all:
$(MAKE) -C src
README.html: README
rst2html $< $@
doc: README.html
rm -rf $(HDDIR)
mkdir -p $(HDDIR)/src
cp hscolour.css $(HDDIR)/src
for file in $(HSRCS); do \
HsColour -css -anchor \
$$file > $(HDDIR)/src/`basename $$file .hs`.html ; \
done
ln -sf hn1.html $(HDDIR)/src/Main.html
haddock --odir $(HDDIR) --html --ignore-all-exports \
-t hn1 -p haddock-prologue \
--source-module="src/%{MODULE/.//}.html" \
--source-entity="src/%{MODULE/.//}.html#%{NAME}" \
$(HSRCS)
clean:
rm -f *.o *.cmi *.cmo *.cmx *.old hn1 zn1 *.prof *.ps *.stat *.aux \
gmon.out *.hi README.html TAGS
.PHONY : all doc clean hn1
Cluster tools (h-aneti?)
========================
These are some simple cluster tools for fixing common problems. Right now N+1
and rebalancing are included.
.. contents::
Cluster N+1 solver
------------------
This program runs a very simple brute force algorithm over the instance
placement space in order to determine the shortest number of replace-disks
needed to fix the cluster. Note this means we won't get a balanced cluster,
just one that passes N+1 checks.
Also note that the set of all instance placements on a 20/80 cluster is
(20*19)^80, that is ~10^200, so...
Algorithm
+++++++++
The algorithm is a simple two-phase process.
In phase 1 we determine the removal set, that is the set of instances that when
removed completely from the cluster, make it healthy again. The instances that
can go into the set are all the primary and secondary instances of the failing
nodes. The result from this phase is actually a list - we compute all sets of
the same minimum length.
So basically we aim to determine here: what is the minimum number of instances
that need to be removed (this is called the removal depth) and which are the
actual combinations that fit (called the list of removal sets).
In phase 2, for each removal set computed in the previous phase, we take the
removed instances and try to determine where we can put them so that the
cluster is still passing N+1 checks. From this list of possible solutions
(called the list of solutions), we compute the one that has the smallest delta
from the original state (the delta is the number of replace disks that needs to
be run) and chose this as the final solution.
Implementation
++++++++++++++
Of course, a naive implementation based on the above description will run for
long periods of time, so the implementation has to be smart in order to prune
the solution space as eagerly as possible.
In the following, we use as example a set of test data (a cluster with 20
nodes, 80 instances that has 5 nodes failing N+1 checks for a total of 12
warnings).
On this set, the minimum depth is 4 (anything below fails), and for this depth
the current version of the algorithm generates 5 removal sets; a previous
version of the first phase generated a slightly different set of instances, with
two removal sets. For the original version of the algorithm:
- the first, non-optimized implementation computed a solution of delta=4 in 30
minutes on server-class CPUs and was still running when aborted 10 minutes
later
- the intermediate optimized version computed the whole solution space and
found a delta=3 solution in around 10 seconds on a laptop-class CPU (total
number of solutions ~600k)
- latest version on server CPUs (which actually computes more removal sets)
computes depth=4 in less than a second and depth=5 in around 2 seconds, and
depth=6 in less than 20 seconds; depth=8 takes under five minutes (this is
10^10 bigger solution space)
Note that when (artificially) increasing the depth to 5 the number of removal
sets grows fast (~3000) and a (again artificial) depth 6 generates 61k removal
sets. Therefore, it is possible to restrict the number of solution sets
examined via a command-line option.
The factors that influence the run time are:
- the removal depth; for each increase with one of the depth, we grow the
solution space by the number of nodes squared (since a new instance can live
any two nodes as primary/secondary, therefore (almost) N times N); i.e.,
depth=1 will create a N^2 solution space, depth two will make this N^4,
depth three will be N^6, etc.
- the removal depth again; for each increase in the depth, there will be more
valid removal sets, and the space of solutions increases linearly with the
number of removal sets
Therefore, the smaller the depth the faster the algorithm will be; it doesn't
seem like this algorithm will work for clusters of 100 nodes and many many
small instances (e.g. 256MB instances on 16GB nodes).
Currently applied optimizations:
- when choosing where to place an instance in phase two, there are N*(N-1)
possible primary/secondary options; however, if instead of iterating over all
p * s pairs, we first determine the set of primary nodes that can hold this
instance (without failing N+1), we can cut (N-1) secondary placements for
each primary node removed; and since this applies at every iteration of phase
2 it linearly decreases the solution space, and on full clusters, this can
mean a four-five times reductions of solution space
- since the number of solutions is very high even for smaller depths (on the
test data, depth=4 results in 1.8M solutions) we can't compare them at the
end, so at each iteration in phase 2 we only promote the best solution out of
our own set of solutions
- since the placement of instances can only increase the delta of the solution
(placing a new instance will add zero or more replace-disks steps), it means
the delta will only increase while recursing during phase 2; therefore, if we
know at one point that we have a current delta that is equal or higher to the
delta of the best solution so far, we can abort the recursion; this cuts a
tremendous number of branches; further promotion of the best solution from
one removal set to another can cut entire removal sets after a few recursions
Command line usage
++++++++++++++++++
Synopsis::
hn1 [-n NODES_FILE] [-i INSTANCES_FILE] [-d START_DEPTH] \
[-r MAX_REMOVALS] [-l MIN_DELTA] [-L MAX_DELTA] \
[-p] [-C]
The -n and -i options change the names of the input files. The -d option
changes the start depth, as a higher depth can give (with a longer computation
time) a solution with better delta. The -r option restricts at each depth the
number of solutions considered - with r=1000 for example even depth=10 finishes
in less than a second.
The -p option will show the cluster state after the solution is implemented,
while the -C option will show the needed gnt-instance commands to implement
it.
The -l (--min-delta) and -L (--max-delta) options restrict the solution in the
following ways:
- min-delta will cause the search to abort early once we find a solution with
delta less than or equal to this parameter; this can cause extremely fast
results in case a desired solution is found quickly; the default value for
this parameter is zero, so once we find a "perfect" solution we finish early
- max-delta causes rejection of valid solution but which have delta higher
than the value of this parameter; this can reduce the depth of the search
tree, with sometimes significant speedups; by default, this optimization is
not used
Individually or combined, these two parameters can (if there are any) very
fast result; on our test data, depth=34 (max depth!) is solved in 2 seconds
with min-delta=0/max-delta=1 (since there is such a solution), and the
extremely low max-delta causes extreme pruning.
Cluster rebalancer
------------------
Compared to the N+1 solver, the rebalancer uses a very simple algorithm:
repeatedly try to move each instance one step, so that the cluster score
becomes better. We stop when no further move can improve the score.
The algorithm is divided into rounds (all identical):
#. Repeat for each instance:
#. Compute score after the potential failover of the instance
#. For each node that is different from the current primary/secondary
#. Compute score after replacing the primary with this new node
#. Compute score after replacing the secondary with this new node
#. Out of this N*2+1 possible new scores (and their associated move) for
this instance, we choose the one that is the best in terms of cluster
score, and then proceed to the next instance
Since we don't compute all combinations of moves for instances (e.g. the first
instance's all moves Cartesian product with second instance's all moves, etc.)
but we proceed serially instance A, then B, then C, the total computations we
make in one steps is simply N(number of nodes)*2+1 times I(number of instances),
instead of (N*2+1)^I. So therefore the runtime for a round is trivial.
Further rounds are done, since the relocation of instances might offer better
places for instances which we didn't move, or simply didn't move to the best
place. It is possible to limit the rounds, but usually the algorithm finishes
after a few rounds by itself.
Note that the cluster *must* be N+1 compliant before this algorithm is run, and
will stay at each move N+1 compliant. Therefore, the final cluster will be N+1
compliant.
Single-round solutions
++++++++++++++++++++++
Single-round solutions have the very nice property that they are
incrementally-valid. In other words, if you have a 10-step solution, at each
step the cluster is both N+1 compliant and better than the previous step.
This means that you can stop at any point and you will have a better cluster.
For this reason, single-round solutions are recommended in the common case of
let's make this better. Multi-round solutions will be better though when adding
a couple of new, empty nodes to the cluster due to the many relocations needed.
Multi-round solutions
+++++++++++++++++++++
A multi-round solution (not for a single round), due to de-duplication of moves
(i.e. just put the instance directly in its final place, and not move it five
times around) loses both these properties. It might be that it's not possible to
directly put the instance on the final nodes. So it can be possible that yes,
the cluster is happy in the final solution and nice, but you cannot do the steps
in the shown order. Solving this (via additional instance move(s)) is left to
the user.
Command line usage
++++++++++++++++++
Synopsis::
hbal [-n NODES_FILE] [-i INSTANCES_FILE] \
[-r MAX_ROUNDS] \
[-p] [-C]
The -n and -i options change the names of the input files. The -r option
restricts the maximum number of rounds (and is more of safety measure).
The -p option will show the cluster state after the solution is implemented,
while the -C option will show the needed gnt-instance commands to implement
it.
Integration with Ganeti
-----------------------
The programs needs only the output of the node list and instance list. That is,
they need the following two commands to be run::
gnt-node list -oname,mtotal,mfree,dtotal,dfree,pinst_list,sinst_list \
--separator '|' --no-headers > nodes
gnt-instance list -oname,admin_ram,sda_size \
--separator '|' --no-head > instances
These two files should be saved under the names of 'nodes' and 'instances'.
When run, the programs will show some informational messages and output the
chosen solution, in the form of a list of instance name and chosen
primary/secondary nodes. The user then needs to run the necessary commands to
get the instances to live on those nodes.
Note that sda_size is less than the total disk size of an instance by 4352
MiB, so if disk space is at a premium the calculation could be wrong; in this
case, please adjust the values manually.
This is the internal documentation for hn1, an experimental N+1
cluster solver.
Start with the "Main" module, the follow with "Cluster" and then the
rest.
.keyglyph, .layout {color: red;}
.keyword {color: blue;}
.comment, .comment a {color: green;}
.str, .chr {color: teal;}
.keyword,.conid, .varid, .conop, .varop, .num, .cpp, .sel, .definition {}
This diff is collapsed.
{-| Module abstracting the node and instance container implementation.
This is currently implemented on top of an 'IntMap', which seems to
give the best performance for our workload.
-}
module Container
(
-- * Types
Container
-- * Creation
, empty
, fromAssocList
-- * Query
, size
, find
-- * Update
, add
, addTwo
, remove
-- * Conversion
, elems
) where
import qualified Data.IntMap as IntMap