Re: RFC: Releasing Ganeti 2.15.3

2016-09-30 Thread 'Iustin Pop' via ganeti-devel
On 30 September 2016 at 12:30, 'Brian Foley' via ganeti-devel <
ganeti-devel@googlegroups.com> wrote:

> On Fri, Sep 30, 2016 at 11:47:24AM +0200, Iustin Pop wrote:
> >On 29 September 2016 at 19:21, 'Brian Foley' via ganeti-devel
> ><[1]ganeti-devel@googlegroups.com> wrote:
> >
> >  Hi all,
> >  since December 2015 quite a large number of commits have been made
> >  to the
> >  stable-2.15 branch (87 to be exact). These cover the gamut from
> >  performance
> >  improvements, to compatibility fixes, to error handling, and even
> >  some minor
> >  feature improvements. There should be no compatibility breaks
> >  though.
> >  You can see the full list with git log --oneline
> >  v2.15.2..stable-2.15
> >  I've tried to categorise them all below. I'd like us to do a
> >  (final?) patch
> >  release of 2.15.3, as the patches includes an important fix for
> >  socat 1.7.3
> >  compatibility that's affected quite a few users.
> >  Does anyone else have anything they'd really like to see fixed in
> >  2.15, or does
> >  anyone have any objections to including any of the below in a point
> >  release?
> >
> >As far as my investigations on
> >[2]https://code.google.com/p/ganeti/issues/detail?id=1185 went, it
> >seems that the location setup tags completely broke htools on clusters
> >with non-DRBD instances. It's a trivial breakage, not a design one, so
> >fixing it shouldn't be too invasive.
> >Day work has been crazy busy so I didn't manage to finish up a patch,
> >but I think this should be fixed in 2.15 stable. When were you
> planning
> >the release for?
>
> Hi Iustin,
>
> I completely forgot about #1185. Yeah, it would definitely be worth fixing
> this if it's not too hard.
>
> I was hoping to get a release done in the next 2-3 weeks if we could, but
> I've no fixed schedule in mind and I'll go with whatever suits people.
>

Ah, OK, that should very doable. I thought you're planning for the next few
days; I'll try to send a patch then sometimes next week.

thanks!
iustin


Re: RFC: Releasing Ganeti 2.15.3

2016-09-30 Thread 'Brian Foley' via ganeti-devel
On Fri, Sep 30, 2016 at 11:47:24AM +0200, Iustin Pop wrote:
>On 29 September 2016 at 19:21, 'Brian Foley' via ganeti-devel
><[1]ganeti-devel@googlegroups.com> wrote:
> 
>  Hi all,
>  since December 2015 quite a large number of commits have been made
>  to the
>  stable-2.15 branch (87 to be exact). These cover the gamut from
>  performance
>  improvements, to compatibility fixes, to error handling, and even
>  some minor
>  feature improvements. There should be no compatibility breaks
>  though.
>  You can see the full list with git log --oneline
>  v2.15.2..stable-2.15
>  I've tried to categorise them all below. I'd like us to do a
>  (final?) patch
>  release of 2.15.3, as the patches includes an important fix for
>  socat 1.7.3
>  compatibility that's affected quite a few users.
>  Does anyone else have anything they'd really like to see fixed in
>  2.15, or does
>  anyone have any objections to including any of the below in a point
>  release?
> 
>As far as my investigations on
>[2]https://code.google.com/p/ganeti/issues/detail?id=1185 went, it
>seems that the location setup tags completely broke htools on clusters
>with non-DRBD instances. It's a trivial breakage, not a design one, so
>fixing it shouldn't be too invasive.
>Day work has been crazy busy so I didn't manage to finish up a patch,
>but I think this should be fixed in 2.15 stable. When were you planning
>the release for?

Hi Iustin,

I completely forgot about #1185. Yeah, it would definitely be worth fixing
this if it's not too hard.

I was hoping to get a release done in the next 2-3 weeks if we could, but
I've no fixed schedule in mind and I'll go with whatever suits people.

Cheers,
Brian.


Re: RFC: Releasing Ganeti 2.15.3

2016-09-30 Thread 'Iustin Pop' via ganeti-devel
On 29 September 2016 at 19:21, 'Brian Foley' via ganeti-devel <
ganeti-devel@googlegroups.com> wrote:

> Hi all,
>
> since December 2015 quite a large number of commits have been made to the
> stable-2.15 branch (87 to be exact). These cover the gamut from performance
> improvements, to compatibility fixes, to error handling, and even some
> minor
> feature improvements. There should be no compatibility breaks though.
>
> You can see the full list with git log --oneline v2.15.2..stable-2.15
>
> I've tried to categorise them all below. I'd like us to do a (final?) patch
> release of 2.15.3, as the patches includes an important fix for socat 1.7.3
> compatibility that's affected quite a few users.
>
> Does anyone else have anything they'd really like to see fixed in 2.15, or
> does
> anyone have any objections to including any of the below in a point
> release?
>

As far as my investigations on
https://code.google.com/p/ganeti/issues/detail?id=1185 went, it seems that
the location setup tags completely broke htools on clusters with non-DRBD
instances. It's a trivial breakage, not a design one, so fixing it
shouldn't be too invasive.

Day work has been crazy busy so I didn't manage to finish up a patch, but I
think this should be fixed in 2.15 stable. When were you planning the
release for?

thanks,
iustin


RFC: Releasing Ganeti 2.15.3

2016-09-29 Thread 'Brian Foley' via ganeti-devel
Hi all,

since December 2015 quite a large number of commits have been made to the
stable-2.15 branch (87 to be exact). These cover the gamut from performance
improvements, to compatibility fixes, to error handling, and even some minor
feature improvements. There should be no compatibility breaks though.

You can see the full list with git log --oneline v2.15.2..stable-2.15

I've tried to categorise them all below. I'd like us to do a (final?) patch
release of 2.15.3, as the patches includes an important fix for socat 1.7.3
compatibility that's affected quite a few users.

Does anyone else have anything they'd really like to see fixed in 2.15, or does
anyone have any objections to including any of the below in a point release?

Appollon, would you be happy with this in 2.15.3 in jessie-backports, sid,
xenial and yakkety? Are there any distro patches we should pull into the
official tree?

Thanks,
Brian Foley.


Compatibility improvements
* d5d747d Use socat method string compatible with <1.73 & >=1.73
* 2429235 Fixup compatibility with GHC 7.4/base 4.5
* 4cea339 568 Update hv_kvm to handle output from qemu >= 1.6.0
* bd7239a KVM: handle gracefully too old/too new psutil versions

Misc
* cc9016f iallocator: only adjust memory usage for up instances
* 203a97f KVM: explicitly configure routed NICs late
* 671d167 mcpu: Raise caught OpPrereqErrors with too few args
* 0df7c24 Bugfix: migrate needs HypervisorClass, not an instance
* 6b6bea0 Fix ganeti-rapi/noded exit-under-load bug
* 8ac8907 Cancel RAPI job if the client drops the connection
* 712ea2f Support userspace disk URIs for OS import/export scripts
* e226032 Give atomicWriteFile temp filenames a more distinct pattern
* 115325f Fix failover in case the source node is offline

Logging/debug improvements
* 2cd22ad Run ssh-key renewal in debug mode during upgrade
* eef3bf4 Also log the high-level upgrade steps
* 249abd0 Add function to provide logged user feedback
* 9d99836 Run renew-crypto in upgrades in debug mode
* 6ee7964 Unconditionally log upgrades at debug level
* 16e95c9 Add debug logging for time to sendMsg in Haskell servers
* 0eaa01f Add debug logging to profile python CallRPCMethod
* f46485a Bracket ConfigWriter writeConfigAndUnlock with debug logging
* 0a89b1d Bracket client LockConfig calls with debug logging
* edfefe7 Get onInotify and onPollTimer to print filepath
* d80f02c Remove extremely frequent debug logging message
* 10e65ed Disable logging CallRPCMethod timings in non-debug configs

Improve master-failover correctness
* bc06ac2 Document healthy-majority restriction on master-failover
* 5115526 Check for healthy majority on master failover with voting
* d2dc0cb Add a predicate testing that a majority of nodes is healthy
* ff009c3 Pass arguments to correct daemons during master-failover

Feature: Allow attaching disks to diskless instances
* 216e20b Test disk attachment with different primary nodes
* 3e55d77 Check for same primary node before disk attachment
* 46ddc97 Add detach/attach sequence test
* 0ea9dd0 Allow disk attachment with external storage

Feature: Rewrite renew-crypto to do O(n) sshes instead of O(n^2)
* bebac87 Catch IOError of SSH files when removing node
* 2e9fdac Fix renew-crypto on one-node-cluster
* b267251 ssh_update: log data that is received
* 61281f9 Increase timeout of RPC adding/removing keys
* 38712d4 After TestNodeModify, fix the pool of master candidates
* 625c8ea Add more documentation to testutils_ssh.py
* cf50a6a renew-crypto: use bulk-removal of SSH keys
* de9087d Use bulk-removal of SSH keys for single keys
* 6fa7813 Bulk-removing SSH keys of diverse set of nodes
* 9ad477c Bulk-removal of SSH keys of normal nodes
* 1699cf0 Bulk-remove SSH keys of potential master candidates
* 125e790 Bulk-removal of SSH keys
* ee428a2 testutils: add keys to own 'authorized_keys' file
* 88610f8 Make mock SSH file manager deal with lists

Misc performance/memory leak fixes
* d262bf7 Don't deepcopy the config if the old value is not needed
* cda51aa Optimise codegen for Python OpCode classes
* 8e4b725 Do not add a new Inotify watchers on timer
* ed1bb06 Optimize LXC hypervisor GetAllInstancesInfo
* 6ea36d2 Fix memory/perf bug in gnt-cluster verify
* f1574de Improve luxid QueryInstances performance for large clusters
* d929e5b Fix ClusterVerifyConfig() causing high mem usage
* 79ee34e Reduce heap when parsing & storing ConfigData 10%
* c429dd2 Make JQScheduler queues more strict to avoid leaks
* cf077d3 Prevent InstanceShutdown from waiting on success

Docs/error handling/comment tweaks
* 2451da1 Fix documentation for master-failover
* 5f6cd34 Fix outdated comment
* 2b51613 Update install-quick DRBD requirements to include DRBD 8.4
* 3b0794a Update misleading comment in noded RPC code
* 1aa3037 Fix CLI option typo in NEWS file
* 9082de6 Make EnsureDirs print chmod mode on errors
* c85311c LV check failure should print instance name

Build-related fixes
* b3c0fb0 Allow newer cabal deps for latest Debian/Ubuntu
* 01e4f28 Fix