Is bcache dead?

Is bcache dead?

[Stefan Priebe - Profihost AG]

Hi,

this might be a bit aggressive - sorry for that.

There are a lot of people having deadlocks with bcache_gc and there's no
dev looking into this since month.

Greets,
Stefan

Re: Is bcache dead?

[Vasiliy Tolstov]

2014-10-30 15:07 GMT+03:00 Stefan Priebe - Profihost AG <s.priebe@profihost.ag>:
> Hi,
>
> this might be a bit aggressive - sorry for that.
>
> There are a lot of people having deadlocks with bcache_gc and there's no
> dev looking into this since month.


As i understand devs provide patches only to next versions of linux.
Now i'm look at dm-cache.... (with lvm frontend lvmcache)

-- 
Vasiliy Tolstov,
e-mail: v.tolstov@selfip.ru
jabber: vase@selfip.ru

Re: Is bcache dead?

[Kent Overstreet]

no, I've just been severely overworked, and overstressed, to the point
that it might be time for a change of jobs - and unfortunately, there
still isn't anyone else who can step in. It's not fun being the single
point of failure.

Sorry for going off the radar. Might be a good time to take some time
off and try and get some upstream work done.

On Thu, Oct 30, 2014 at 5:07 AM, Stefan Priebe - Profihost AG
<s.priebe@profihost.ag> wrote:
> Hi,
>
> this might be a bit aggressive - sorry for that.
>
> There are a lot of people having deadlocks with bcache_gc and there's no
> dev looking into this since month.
>
> Greets,
> Stefan

Re: Is bcache dead?

[Stefan Priebe - Profihost AG]

Am 30.10.2014 um 15:09 schrieb Vasiliy Tolstov:
> 2014-10-30 15:07 GMT+03:00 Stefan Priebe - Profihost AG <s.priebe@profihost.ag>:
>> Hi,
>>
>> this might be a bit aggressive - sorry for that.
>>
>> There are a lot of people having deadlocks with bcache_gc and there's no
>> dev looking into this since month.
> 
> 
> As i understand devs provide patches only to next versions of linux.
> Now i'm look at dm-cache.... (with lvm frontend lvmcache)

no problem but this bug exists since 3.15 and is not fixed in 3.18-rc..

Stefan

Re: Is bcache dead?

[Kent Overstreet]

since 3.15!? shit, I don't think I saw any reports until 3.17 came out...

On Thu, Oct 30, 2014 at 7:14 AM, Stefan Priebe - Profihost AG
<s.priebe@profihost.ag> wrote:
>
> Am 30.10.2014 um 15:09 schrieb Vasiliy Tolstov:
>> 2014-10-30 15:07 GMT+03:00 Stefan Priebe - Profihost AG <s.priebe@profihost.ag>:
>>> Hi,
>>>
>>> this might be a bit aggressive - sorry for that.
>>>
>>> There are a lot of people having deadlocks with bcache_gc and there's no
>>> dev looking into this since month.
>>
>>
>> As i understand devs provide patches only to next versions of linux.
>> Now i'm look at dm-cache.... (with lvm frontend lvmcache)
>
> no problem but this bug exists since 3.15 and is not fixed in 3.18-rc..
>
> Stefan
>

Re: Is bcache dead?

[Stefan Priebe - Profihost AG]

Hi Kent,

nice to hear you. No problem at all. I was just wondering that there was
no single response to all the bcache_gc problems people have,

Greets,
Stefan

Am 30.10.2014 um 15:14 schrieb Kent Overstreet:
> no, I've just been severely overworked, and overstressed, to the point
> that it might be time for a change of jobs - and unfortunately, there
> still isn't anyone else who can step in. It's not fun being the single
> point of failure.
> 
> Sorry for going off the radar. Might be a good time to take some time
> off and try and get some upstream work done.
> 
> On Thu, Oct 30, 2014 at 5:07 AM, Stefan Priebe - Profihost AG
> <s.priebe@profihost.ag> wrote:
>> Hi,
>>
>> this might be a bit aggressive - sorry for that.
>>
>> There are a lot of people having deadlocks with bcache_gc and there's no
>> dev looking into this since month.
>>
>> Greets,
>> Stefan

Re: Is bcache dead?

[Stefan Priebe - Profihost AG]

Am 30.10.2014 um 15:15 schrieb Kent Overstreet:
> since 3.15!? shit, I don't think I saw any reports until 3.17 came out...

maybe i'm wrong. Can't guarantee but i already saw it on 3.16.X too.

> On Thu, Oct 30, 2014 at 7:14 AM, Stefan Priebe - Profihost AG
> <s.priebe@profihost.ag> wrote:
>>
>> Am 30.10.2014 um 15:09 schrieb Vasiliy Tolstov:
>>> 2014-10-30 15:07 GMT+03:00 Stefan Priebe - Profihost AG <s.priebe@profihost.ag>:
>>>> Hi,
>>>>
>>>> this might be a bit aggressive - sorry for that.
>>>>
>>>> There are a lot of people having deadlocks with bcache_gc and there's no
>>>> dev looking into this since month.
>>>
>>>
>>> As i understand devs provide patches only to next versions of linux.
>>> Now i'm look at dm-cache.... (with lvm frontend lvmcache)
>>
>> no problem but this bug exists since 3.15 and is not fixed in 3.18-rc..
>>
>> Stefan
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: Is bcache dead?

[Thomas Stein]

On Thursday 30 October 2014 15:18:54 Stefan Priebe - Profihost AG wrote:
> Am 30.10.2014 um 15:15 schrieb Kent Overstreet:
> > since 3.15!? shit, I don't think I saw any reports until 3.17 came out...
> 
> maybe i'm wrong. Can't guarantee but i already saw it on 3.16.X too.

We should not forget the "load average not falling bellow 1" bug which is 
still present in 3.17.1.

cheers
t.

> > On Thu, Oct 30, 2014 at 7:14 AM, Stefan Priebe - Profihost AG
> > 
> > <s.priebe@profihost.ag> wrote:
> >> Am 30.10.2014 um 15:09 schrieb Vasiliy Tolstov:
> >>> 2014-10-30 15:07 GMT+03:00 Stefan Priebe - Profihost AG 
<s.priebe@profihost.ag>:
> >>>> Hi,
> >>>> 
> >>>> this might be a bit aggressive - sorry for that.
> >>>> 
> >>>> There are a lot of people having deadlocks with bcache_gc and there's
> >>>> no
> >>>> dev looking into this since month.
> >>> 
> >>> As i understand devs provide patches only to next versions of linux.
> >>> Now i'm look at dm-cache.... (with lvm frontend lvmcache)
> >> 
> >> no problem but this bug exists since 3.15 and is not fixed in 3.18-rc..
> >> 
> >> Stefan
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Is bcache dead?

[Stan Hoeppner]

On 10/30/2014 09:14 AM, Kent Overstreet wrote:
> no, I've just been severely overworked, and overstressed, to the point
> that it might be time for a change of jobs - and unfortunately, there
> still isn't anyone else who can step in. It's not fun being the single
> point of failure.

Don't sweat it Kent.  Don't get discouraged.  Stay positive.

I tried bcache a few weeks ago for a pretty niche application and it
wasn't suitable for that workload for what I wanted it to do.  I asked
questions here to make it work but got no responses.  Would have been a
feather in your cap to had bcache on those systems--two 44TiB LUNs on
the small ones, 14x 44TiB LUNs on the large one--if it could have been
made to work with that workload.  We'll probably fix it by modifying the
app to do full stripe buffer writes.  Yes, this is much more work than
simply slapping in bcache, had it worked.  I was looking for a quick fix.

I know the demands from myself and others can create stress.  But when
you feel stressed by it, remember that the demand is a direct result of
you creating something special, that people really want to use.

You recognize and acknowledge the fact that you're a one man show right
now.  Your users know it too.  Do what you can when you can, and do it
right.  I think most people will be more forgiving of delays than
mistakes, or broken promises, or silence.  Communication helps.  If
you're bogged down, just post a quick note the list letting everyone
know.  A quick update like that goes a long way, whereas silence breeds
discontent among users, because they don't know what's going on.

Keep your chin up.  You'll get there, even if it takes longer than folks
would like.

Best regards,

Stan

Re: Is bcache dead?

[Kent Overstreet]

On Thu, Oct 30, 2014 at 02:55:03PM -0500, Stan Hoeppner wrote:
> On 10/30/2014 09:14 AM, Kent Overstreet wrote:
> > no, I've just been severely overworked, and overstressed, to the point
> > that it might be time for a change of jobs - and unfortunately, there
> > still isn't anyone else who can step in. It's not fun being the single
> > point of failure.
> 
> Don't sweat it Kent.  Don't get discouraged.  Stay positive.
> 
> I tried bcache a few weeks ago for a pretty niche application and it
> wasn't suitable for that workload for what I wanted it to do.  I asked
> questions here to make it work but got no responses.  Would have been a
> feather in your cap to had bcache on those systems--two 44TiB LUNs on
> the small ones, 14x 44TiB LUNs on the large one--if it could have been
> made to work with that workload.  We'll probably fix it by modifying the
> app to do full stripe buffer writes.  Yes, this is much more work than
> simply slapping in bcache, had it worked.  I was looking for a quick fix.
> 
> I know the demands from myself and others can create stress.  But when
> you feel stressed by it, remember that the demand is a direct result of
> you creating something special, that people really want to use.
> 
> You recognize and acknowledge the fact that you're a one man show right
> now.  Your users know it too.  Do what you can when you can, and do it
> right.  I think most people will be more forgiving of delays than
> mistakes, or broken promises, or silence.  Communication helps.  If
> you're bogged down, just post a quick note the list letting everyone
> know.  A quick update like that goes a long way, whereas silence breeds
> discontent among users, because they don't know what's going on.
> 
> Keep your chin up.  You'll get there, even if it takes longer than folks
> would like.
> 
> Best regards,
> 
> Stan

Thanks, I really do appreciate the kind words.

Re: Is bcache dead?

[Zachary Palmer]

> On Thu, Oct 30, 2014 at 02:55:03PM -0500, Stan Hoeppner wrote:
>> On 10/30/2014 09:14 AM, Kent Overstreet wrote:
>>> no, I've just been severely overworked, and overstressed, to the point
>>> that it might be time for a change of jobs - and unfortunately, there
>>> still isn't anyone else who can step in. It's not fun being the single
>>> point of failure.
>> Don't sweat it Kent.  Don't get discouraged.  Stay positive.
>>
>> I tried bcache a few weeks ago for a pretty niche application and it
>> wasn't suitable for that workload for what I wanted it to do.  I asked
>> questions here to make it work but got no responses.  Would have been a
>> feather in your cap to had bcache on those systems--two 44TiB LUNs on
>> the small ones, 14x 44TiB LUNs on the large one--if it could have been
>> made to work with that workload.  We'll probably fix it by modifying the
>> app to do full stripe buffer writes.  Yes, this is much more work than
>> simply slapping in bcache, had it worked.  I was looking for a quick fix.
>>
>> I know the demands from myself and others can create stress.  But when
>> you feel stressed by it, remember that the demand is a direct result of
>> you creating something special, that people really want to use.
>>
>> You recognize and acknowledge the fact that you're a one man show right
>> now.  Your users know it too.  Do what you can when you can, and do it
>> right.  I think most people will be more forgiving of delays than
>> mistakes, or broken promises, or silence.  Communication helps.  If
>> you're bogged down, just post a quick note the list letting everyone
>> know.  A quick update like that goes a long way, whereas silence breeds
>> discontent among users, because they don't know what's going on.
>>
>> Keep your chin up.  You'll get there, even if it takes longer than folks
>> would like.
>>
>> Best regards,
>>
>> Stan
> Thanks, I really do appreciate the kind words.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
If I can throw mine in as well, I've been running bcache on my Debian 
Wheezy laptop for around a year now.  (I'm using the Debian backports 
3.12 kernel.)  When I first moved to bcache, I noticed that certain 
operations -- interacting with Git repositories and building LaTeX 
documents, for instance -- became much snappier.  I'm using a feeble 
little 32GB SSD that came with the laptop to cache a 1TB drive and I'm 
even using writethrough caching (more out of paranoia about the quality 
of my cheap little SSD than anything else), but it makes a difference.

Since then, it has been quietly humming along and I've stopped paying 
attention to it.  And that's the beauty of a good tool like this: I can 
stop paying attention to it.  I've enjoyed a year of better I/O and, 
other than in the initial setup, I haven't paid anything in maintenance 
burden: no instability, no hiccups, no unexplained hangs.  So for my 
part as an end user just trying to get a little edge out of my laptop 
hardware, thank you!  I expect I'm speaking on behalf of quite a few 
people when I say that you've made things better in a subtle but 
significant way.  :)

Cheers,

Zach

Re: Is bcache dead?

[Rolf Fokkens]

Another happy user experience: I started packaging bcache-tools for 
Fedora 20 summer 2013, just because I wanted to use bcache. After 
alligning bcache-tools with util-linux, Dracut and LVM2 (all these 
packages needed minor tweaks, that were all integrated upstream) I have 
been using bcache myself (of course). I've been living on the edge by 
using writeback caching and a cheap SSD, and it has all been working 
like a charm! Well, I've been living over the edge actually by 
attempting to enable TRIM - I blame the resulting corruptions on the 
cheap SSD. And I had occasional "bcache_writeback 100% CPU" issues, but 
those seem to be gone (currently kernel 3.16).

And searching for bcache at bugzilla.redhat.com: no bugs pop up. Could 
be that there are no bcache users at all, but I know for fact that 
that's not true.

So (when not using TRIM): excellent performance and stability. Thanks 
for the good work!

Rolf

On 11/06/2014 04:03 PM, Zachary Palmer wrote:
> If I can throw mine in as well, I've been running bcache on my Debian 
> Wheezy laptop for around a year now.  (I'm using the Debian backports 
> 3.12 kernel.)  When I first moved to bcache, I noticed that certain 
> operations -- interacting with Git repositories and building LaTeX 
> documents, for instance -- became much snappier. I'm using a feeble 
> little 32GB SSD that came with the laptop to cache a 1TB drive and I'm 
> even using writethrough caching (more out of paranoia about the 
> quality of my cheap little SSD than anything else), but it makes a 
> difference.
>
> Since then, it has been quietly humming along and I've stopped paying 
> attention to it.  And that's the beauty of a good tool like this: I 
> can stop paying attention to it.  I've enjoyed a year of better I/O 
> and, other than in the initial setup, I haven't paid anything in 
> maintenance burden: no instability, no hiccups, no unexplained hangs.  
> So for my part as an end user just trying to get a little edge out of 
> my laptop hardware, thank you!  I expect I'm speaking on behalf of 
> quite a few people when I say that you've made things better in a 
> subtle but significant way.  :)
>
> Cheers,
>
> Zach

Re: Is bcache dead?

[Michael Russo]

Zachary Palmer <zep_kernel.org <at> bahj.com> writes:

>  I expect I'm speaking on behalf of quite a few 
> people when I say that you've made things better in a subtle but 
> significant way.  :)
> 
> Cheers,
> 
> Zach
> 

I am a little late in this thread but I also wanted to say I appreciate the hard
 work put into bcache and I've had no trouble with it at all, except when my 
crappy OCZ VERTEX4 disappears (happens maybe once a month, or if the case gets a
 little too hot inside).  Then I'll need to reboot whether I'm in writeback 
OR writethrough mode (I've got bcache as my root FS and in both cases it's 
remounted r/o and bcache won't let me remount it r/w).  But it really makes 
things  faster (in either mode, though of course faster in writeback). Hopefully
 the stress level will go down and you will get some more time to work on 
bcache.