root / NEWS @ b7b29191
History | View | Annotate | Download (16.1 kB)
1 |
Ganeti-htools release notes |
---|---|
2 |
=========================== |
3 |
|
4 |
|
5 |
Version 0.2.7 (Thu, 07 Oct 2010) |
6 |
-------------------------------- |
7 |
|
8 |
Bug fixes: |
9 |
|
10 |
- fixed the error message for hail multi-evacuation mode |
11 |
- improve evacuation mode for offline secondary nodes (ignore available |
12 |
memory) |
13 |
|
14 |
New features: |
15 |
|
16 |
- add a new option ``-S`` to hbal and hspace that saves the cluster |
17 |
state at the end of the processing in the text format used by the |
18 |
``-t`` option, for later re-processing |
19 |
- a two new options to hbal, -g and --min-gain-limit, that should help |
20 |
in limiting the number of balances steps with a low gain in the final |
21 |
stages |
22 |
- hbal, when executing jobs, will now wait for the current jobs to |
23 |
finish at the first stop (e.g. ^C); if the user wants immediate exit, |
24 |
another signal should be sent |
25 |
- added “normalized” physical CPU units in hspace output (NPU), which |
26 |
represents units of physical CPUs free/used, based on the max-cpu |
27 |
ratio |
28 |
|
29 |
|
30 |
Version 0.2.6 (Mon, 26 Jul 2010) |
31 |
-------------------------------- |
32 |
|
33 |
Exactly three months since the last release. Many internal changes, plus |
34 |
a couple of important changes in the balancing algorithm. |
35 |
|
36 |
First, the balancing may now introduce N+1 errors, if this solves other, |
37 |
more critical problems. For the moment, this means that moving instances |
38 |
away from offline nodes is allowed even if it creates N+1 errors, and |
39 |
that means evacuation can be done in more cases. |
40 |
|
41 |
Second, the scoring for N+1 has changed. In previous versions, it simply |
42 |
counted the number of failing N+1 nodes, which means moving an instance |
43 |
away from a N+1 failed node (but without the node 'clearing' the N+1 |
44 |
status) was not reflected in the cluster score. As such, the balancing |
45 |
algorithm managed to clear N+1 errors only sometimes, since usually it |
46 |
takes more than one move for this, and the first prerequisite move was |
47 |
not 'rewarded' appropriately and thus it was not selected. Now, it is |
48 |
possible to fix many more error cases than before: on a simulated 40 |
49 |
node cluster full with instances (symmetrically allocated on all nodes), |
50 |
around five nodes can be evacuated before N+1 errors can be solved, |
51 |
whereas 0.2.5 could evacuate at best one node. |
52 |
|
53 |
There were some other internal changes to the scoring algorithm, such |
54 |
that now the metrics have associated weights, and they are not all of |
55 |
the same importance anymore. As of now, the only change is that offline |
56 |
instances have a higher weight, which should favour proper node |
57 |
evacuations. |
58 |
|
59 |
Among the other changes: |
60 |
|
61 |
- fixed the hspace KM_POOL_* metrics, which were returned as the final |
62 |
state and not as the delta between the initial and final states |
63 |
- fixed hspace handling of N+1 failing clusters: before, it used to |
64 |
generate a 'fake' response, and the structure of this response was not |
65 |
always in sync with the real responses, leading to missing items; |
66 |
currently it proceeds correctly through the code (skipping the |
67 |
computation), and uses the same display mechanisms as the normal case |
68 |
- fixed hscan exit code for RAPI failures: previously it finished with |
69 |
success even if all the clusters failed, which was creating issues |
70 |
with the live-test script; now it exits with exit code 2 for RAPI |
71 |
failures (unfortunately this is still not optimal as LUXI failures |
72 |
will use exit code 1, the same as the command line) |
73 |
- changed the limit values for CPU/disk, which previously were used |
74 |
optionally, whereas now they are always used; the default cpu ratio |
75 |
limit is now 64 VCPUs per PCPU |
76 |
- changed the internal handling of the short name vs. original |
77 |
(Ganeti-provided) name; now internally we always use the full name, |
78 |
and only in display routines we show the shortened (called 'alias') |
79 |
name; as a result, the -O and --excluded-instances options now accept |
80 |
both the full name and the shortened name |
81 |
- changed internal handling of JSON conversions and errors, such that |
82 |
now we show a better context for failure messages, which should help |
83 |
with diagnosing the malformed message |
84 |
- changed the names for a few node fields, and added some more nodes; |
85 |
this is most likely to help with debugging, and not with regular |
86 |
operation though |
87 |
- changed the node fields option to allow the '+' prefix to mean 'extend |
88 |
the default fields list' rather than start from fresh (similar to |
89 |
Ganeti's implementation) |
90 |
- a few internal changes related to the LUXI protocol implementation, |
91 |
which should make it more safe against potential bugs, one |
92 |
optiomization that should help with large messages, and some patches |
93 |
in preparation for potential expansion of the LUXI backend functionality |
94 |
|
95 |
And finally, many improvements on unittests and the live-test |
96 |
script. Test coverage is much enhanced, and the test infrastructure has |
97 |
better error reporting; this should lead down-the-road to better code |
98 |
and fewer bugs… |
99 |
|
100 |
|
101 |
Version 0.2.5 (Mon, 26 Apr 2010) |
102 |
-------------------------------- |
103 |
|
104 |
Some internal cleanup plus a few user-visible changes: |
105 |
|
106 |
- new option for marking instances as 'do-not-move' during rebalancing |
107 |
- allow ``hscan`` to scan the local cluster via Luxi |
108 |
- add more metrics to ``hspace`` which show the delta between original |
109 |
state and final state better (only valid for tiered allocation) |
110 |
|
111 |
|
112 |
Version 0.2.4 (Mon, 22 Feb 2010) |
113 |
-------------------------------- |
114 |
|
115 |
Two improvements for node evacuation: |
116 |
|
117 |
- hbal takes a new parameter ``--evac-mode`` that restricts the |
118 |
instances to be moved to the ones on offline/drained nodes, which |
119 |
should reduce the work done |
120 |
- hail supports the new ``multi-evacuate`` mode of the IAllocator |
121 |
protocol, that will be released in a minor release on the Ganeti 2.1 |
122 |
branch |
123 |
|
124 |
|
125 |
Version 0.2.3 (Thu, 4 Feb 2010) |
126 |
-------------------------------- |
127 |
|
128 |
A small release: |
129 |
|
130 |
- Fixes selection of secondary node: previously, if the cluster had |
131 |
many N+1 failures, a N+1 failed node could be selected as secondary |
132 |
even if it did not have enough memory to allow the instance to be |
133 |
migrated/failed over to it; this is bad for automated tools, since |
134 |
we can get the cluster in an unhealthy state |
135 |
- Switch the text backend to a single input file, that is generated |
136 |
now by hscan and shouldn't be generated manually via |
137 |
gnt-node/instance list anymore; this allows richer information to be |
138 |
kept in the file, and simplifies a little the internals of the text |
139 |
backend |
140 |
|
141 |
|
142 |
Version 0.2.2 (Tue, 29 Dec 2009) |
143 |
-------------------------------- |
144 |
|
145 |
Small release, 0.2.1 was broken and thus this was released earlier: |
146 |
|
147 |
- Release 0.2.1 broke the LUXI backend due to a typo, fixed |
148 |
- Added a live-test script that should catch errors like the above one |
149 |
in the future (needs a working, non-empty cluster) |
150 |
- Changed RAPI and LUXI backends to treat drained nodes as offline, |
151 |
similar to the IAllocator backend change in 0.2.0 (which was wrongly |
152 |
marked as affecting all backends) |
153 |
- Changed the metrics for offline instances and N1 score from percent to |
154 |
count, in order to increase the priority of evacuations |
155 |
- Added a new metric (offline primary instances) which should fix the |
156 |
evacuation of a offline node in a 2-node cluster |
157 |
|
158 |
|
159 |
Version 0.2.1 (Wed, 2 Dec 2009) |
160 |
-------------------------------- |
161 |
|
162 |
- Added instance exclusion defined via instance tags |
163 |
- Fixed the output of hspace to be again parseable from the shell |
164 |
|
165 |
|
166 |
Version 0.2.0 (Tue, 10 Nov 2009) |
167 |
-------------------------------- |
168 |
|
169 |
A significant release, with a few new major features: |
170 |
|
171 |
- Added direct execution of the hbal solution when using the Luxi |
172 |
backend; the steps for each instance moves are submitted as a single |
173 |
jobs, and the different jobs are submitted as groups in order to |
174 |
parallelise the execution of moves |
175 |
- Added support for balancing based on dynamic utilisation data for |
176 |
instances, fed in via a text file; by default, all instances are |
177 |
considered equal and this change also improves the equalisation of |
178 |
secondary instances per node |
179 |
- Added support for tiered capacity calculation in hspace, where we |
180 |
start from a maximum instance spec and decrease the spec when we run |
181 |
out of resources; this should give a better measure of available |
182 |
capacity on 'fragmented' clusters; this is done separately from the |
183 |
current fixed-mode computation |
184 |
|
185 |
Also there have been many minor improvements: |
186 |
|
187 |
- Added option for showing instances (“--print-instances”), similar to |
188 |
the print nodes option |
189 |
- Added support for customising the node list via an argument to the |
190 |
print nodes option in the form of a comma-separated list of field |
191 |
names; currently the field names are not documented, expecting further |
192 |
changes in a next release |
193 |
- Enhanced the error reporting in the Luxi and Rapi backends |
194 |
- Changed the handling of drained nodes, now being treated the same as |
195 |
offline nodes, for Ganeti 2.0.4+ compatibility |
196 |
- A number of internal changes, simplifying code and merging some |
197 |
disparate functions |
198 |
- Simplify the build system in relation to creation of archives |
199 |
|
200 |
|
201 |
Version 0.1.8 (Tue, 29 Sep 2009) |
202 |
-------------------------------- |
203 |
|
204 |
- Brown-paper-bag release fixing haddock issues |
205 |
|
206 |
|
207 |
Version 0.1.7 (Mon, 28 Sep 2009) |
208 |
-------------------------------- |
209 |
|
210 |
- Fixed a bug in the Luxi backend for big responses |
211 |
- Fixed test suite exit code in presence of test failures |
212 |
- Changed the migrate operation to run instead failover for instances |
213 |
which were marked as not running in the input data (this could have |
214 |
been changed since then, but it's better than today's always migrate) |
215 |
- Added support for 'cheap' moves only (only migrate/failover) in |
216 |
balancing |
217 |
- Added support for building without curl (thus no RAPI backend) |
218 |
|
219 |
|
220 |
Version 0.1.6 (Wed, 19 Aug 2009) |
221 |
-------------------------------- |
222 |
|
223 |
- Added support for Luxi (the native Ganeti protocol) |
224 |
- Added support for simulated clusters (for hspace only) |
225 |
- Added timeouts for the RAPI backend |
226 |
- Fixed a few inconsistencies in the command line handling |
227 |
- Fixed handling of errors while loading data |
228 |
- The 'network' is a new dependency due to the Luxi addition |
229 |
|
230 |
|
231 |
Version 0.1.5 (Thu, 09 Jul 2009) |
232 |
-------------------------------- |
233 |
|
234 |
- Removed obsolete hn1 program; this allowed removal of a lot of |
235 |
supporting code |
236 |
- Lots of changes in hspace: the output now is a shell fragment in order |
237 |
for script to source it or parse it easier; added failure reasons; |
238 |
optimised to use less memory for large clusters |
239 |
- Optimized the scoring algorithm (used by all tools) so that now |
240 |
computations should be faster |
241 |
|
242 |
|
243 |
Version 0.1.4 (Tue, 16 Jun 2009) |
244 |
-------------------------------- |
245 |
|
246 |
- Added CPU count/ratio of virtual-to-physical CPUs to the cluster |
247 |
scoring methods; this means that now the balancer, the iallocator |
248 |
plugin and so on will try to keep the VCPU-to-PCPU ratio equal across |
249 |
the cluster |
250 |
- Fixed some hscan bugs |
251 |
- Fixed the way iallocator reads the total disk size (was broken and it |
252 |
was always falling back to summing the disk sizes) |
253 |
- Internals: fixed most compile-time warnings |
254 |
|
255 |
|
256 |
Version 0.1.3 (Fri, 05 Jun 2009) |
257 |
-------------------------------- |
258 |
|
259 |
- Fix a bug in the ReplacePrimary instance moves, affecting most of the |
260 |
tools |
261 |
|
262 |
|
263 |
Version 0.1.2 (Tue, 02 Jun 2009) |
264 |
-------------------------------- |
265 |
|
266 |
- Add a new program, “hspace”, which computes the free space on a |
267 |
cluster (based on a given instance spec) |
268 |
- Improvements in API docs and partially in the user docs |
269 |
- Started adding unittests |
270 |
|
271 |
|
272 |
Version 0.1.1 (Tue, 26 May 2009) |
273 |
-------------------------------- |
274 |
|
275 |
- Add a new program, “hail”, which is an iallocator plugin and can |
276 |
allocate/relocate instances |
277 |
- Experimental support for non-mirrored instances (hail supports them, |
278 |
hbal should no longer abort when it finds such instances and simply |
279 |
ignore them) |
280 |
- The RAPI port and/or scheme can be overriden now, and even “file://” |
281 |
schemes can be used if the message body has been saved under the |
282 |
appropriate name |
283 |
- Lots of code reorganization, esp. rewritten loading pipeline |
284 |
- Better data checking and better error messages in case validation |
285 |
fails; tools now consider nodes with error in input data (‘?’ returned |
286 |
by ganeti) as offline |
287 |
- Small enhancement to the makefile for simpler packaging |
288 |
|
289 |
|
290 |
Version 0.1.0 (Tue, 19 May 2009) |
291 |
-------------------------------- |
292 |
|
293 |
- Drop compatibility with Ganeti 1.2 |
294 |
- Add a new minimum score option (with a very low default), should help |
295 |
with very good clusters (but is still not optimal) |
296 |
- Add a --quiet option to hbal |
297 |
- Add support for reading offline nodes directly from the cluster |
298 |
|
299 |
|
300 |
Version 0.0.8 (Tue, 21 Apr 2009) |
301 |
-------------------------------- |
302 |
|
303 |
- hbal: prevent mismatches in wrong node names being passed to -O, by |
304 |
aborting in this case |
305 |
- add the ability to write the commands (-C) to a script via (-C<file>), |
306 |
so that it can be later executed directly; this has also changed the |
307 |
commands to include the ncessary -f flags to skip confirmations |
308 |
- add checks for extra argument in hbal and hn1, so that unintended |
309 |
errors are catched |
310 |
- raise the accepted “missing” memory limit to 512MB, to cover usual Xen |
311 |
reservations |
312 |
|
313 |
|
314 |
Version 0.0.7 (Mon, 23 Mar 2009) |
315 |
-------------------------------- |
316 |
|
317 |
- added support for offline nodes, which are not used as targets for |
318 |
instance relocation and if they hold instances the hbal algorithm will |
319 |
attempt to relocate these away |
320 |
- added support for offline instances, which now will no longer skew the |
321 |
free memory estimation of nodes; the algorithm will no longer create |
322 |
conditions for N+1 failures when such instances are later started |
323 |
- implemented a complete model of node resources, in order to prevent an |
324 |
unintended re-occurrence of cases like the offline instance were we |
325 |
miscalculate some node resource; this gives warning now in case the |
326 |
node reported free disk or free memory deviates by more than a set |
327 |
amount from the expected value |
328 |
- a new tool *hscan* that can generate the input text-file for the other |
329 |
tools by collection via RAPI |
330 |
- some small changes to the build system to make it more friendly; also |
331 |
included the generated documentation in the source archive |
332 |
|
333 |
|
334 |
Version 0.0.6 (Mon, 16 Mar 2009) |
335 |
-------------------------------- |
336 |
|
337 |
- re-factored the hbal algorithm to make it stable in the sense that it |
338 |
gives the same solution when restarted from the middle; barring |
339 |
rounding of disk/memory and incomplete reporting from Ganeti (for |
340 |
1.2), it should be now feasible to rely on its output without |
341 |
generating moves ad infinitum |
342 |
- the hbal algorithm now uses two more variables: the node N+1 failures |
343 |
and the amount of reserved memory; the first of which tries to ‘fix’ |
344 |
the N+1 status, the latter tries to distribute secondaries more |
345 |
equally |
346 |
- the hbal algorithm now uses two more moves at each step: |
347 |
replace+failover and failover+replace (besides the original failover, |
348 |
replace, and failover+replace+failover) |
349 |
- slightly changed the build system to embed GIT version/tags into the |
350 |
binaries so that we know for a binary from which tree it was done, |
351 |
either via ‘--version’ or via “strings hbal|grep version” |
352 |
- changed the solution list and in general the hbal output to be more |
353 |
clear by default, and changed “gnt-instance failover” to “gnt-instance |
354 |
migrate” |
355 |
- added man pages for the two binaries |
356 |
|
357 |
|
358 |
Version 0.0.5 (Mon, 09 Mar 2009) |
359 |
-------------------------------- |
360 |
|
361 |
- a few small improvements for hbal (possibly undone by later changes), |
362 |
hbal is now quite faster |
363 |
- fix documentation building |
364 |
- allow hbal to work on non N+1 compliant clusters, but without |
365 |
guarantees that the end cluster will be compliant; in any case, this |
366 |
should give a smaller number of nodes that are not compliant if the |
367 |
cluster state permits it |
368 |
- strip common domain suffix from nodes and instances, so that output is |
369 |
shorter and hopefully clearer |
370 |
|
371 |
|
372 |
Version 0.0.4 (Sun, 15 Feb 2009) |
373 |
-------------------------------- |
374 |
|
375 |
- better balancing algorithm in hbal |
376 |
- implemented an RAPI collector, now the cluster data can be gathered |
377 |
automatically via RAPI and doesn't need manual export of node and |
378 |
instance list |
379 |
|
380 |
|
381 |
Version 0.0.3 (Wed, 28 Jan 2009) |
382 |
-------------------------------- |
383 |
|
384 |
- initial release of the hbal, a cluster rebalancing tool |
385 |
- input data format changed due to hbal requirements |
386 |
|
387 |
|
388 |
Version 0.0.2 (Tue, 06 Jan 2009) |
389 |
-------------------------------- |
390 |
|
391 |
- fix handling of some common cases (cluster N+1 compliant from the |
392 |
start, too big depth given, failure to compute solution) |
393 |
- add option to print the needed command list for reaching the proposed |
394 |
solution |
395 |
|
396 |
|
397 |
Version 0.0.1 (Tue, 06 Jan 2009) |
398 |
-------------------------------- |
399 |
|
400 |
- initial release of hn1 tool |
401 |
|
402 |
.. vim: set textwidth=72 : |
403 |
.. Local Variables: |
404 |
.. mode: rst |
405 |
.. fill-column: 72 |
406 |
.. End: |