Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
MIME-Version: 1.0
In-Reply-To: <etPan.5914b459.1a8f073d.e21@apple.com>
References: <CAKy_EwZ=eVAkDwBY12Vcw4chjjp7=0hiK4RwT2t_7SU5E_9SGA@mail.gmail.com>
 <etPan.5914b459.1a8f073d.e21@apple.com>
From: Stefano Ortolani <ostefano@gmail.com>
Date: Fri, 12 May 2017 14:22:30 +0100
Message-ID: <CAKy_EwaLF108BX1EVpjJmqVZQ183eDP-nFDZNqN2PzdEguUjkQ@mail.gmail.com>
Subject: Re: LCS, range tombstones, and eviction
To: Blake Eggleston <beggleston@apple.com>
Cc: user@cassandra.apache.org
Content-Type: multipart/alternative; boundary="001a113e07d8237031054f539ac5"
archived-at: Fri, 12 May 2017 13:22:58 -0000

--001a113e07d8237031054f539ac5
Content-Type: text/plain; charset="UTF-8"

Thanks a lot Blake, that definitely helps!

I actually found a ticket re range tombstones and how they are accounted
for: https://issues.apache.org/jira/browse/CASSANDRA-8527

I am wondering now what happens when a node receives a read request. Are
the range tombstones read before scanning the SStables? More interestingly,
given that a single partition might be split across different levels, and
that some range tombstones might be in L0 while all the rest of the data in
L1, are all the tombstones prefetched from _all_ the involved SStables
before doing any table scan?

Regards,
Stefano

On Thu, May 11, 2017 at 7:58 PM, Blake Eggleston <beggleston@apple.com>
wrote:

> Hi Stefano,
>
> Based on what I understood reading the docs, if the ratio of garbage
> collectable tomstones exceeds the "tombstone_threshold", C* should start
> compacting and evicting.
>
>
> If there are no other normal compaction tasks to be run, LCS will attempt
> to compact the sstables it estimates it will be able to drop the most
> tombstones from. It does this by estimating the number of tombstones an
> sstable has that have passed the gc grace period. Whether or not a
> tombstone will actually be evicted is more complicated. Even if a tombstone
> has passed gc grace, it can't be dropped if the data it's deleting still
> exists in another sstable, otherwise the data would appear to return. So, a
> tombstone won't be dropped if there is data for the same partition in other
> sstables that is older than the tombstone being evaluated for eviction.
>
> I am quite puzzled however by what might happen when dealing with range
> tombstones. In that case a single tombstone might actually stand for an
> arbitrary number of normal tombstones. In other words, do range tombstones
> contribute to the "tombstone_threshold"? If so, how?
>
>
> From what I can tell, each end of the range tombstone is counted as a
> single tombstone tombstone. So a range tombstone effectively contributes
> '2' to the count of tombstones for an sstable. I'm not 100% sure, but I
> haven't seen any sstable writing logic that tracks open tombstones and
> counts covered cells as tombstones. So, it's likely that the effect of
> range tombstones covering many rows are under represented in the droppable
> tombstone estimate.
>
> I am also a bit confused by the "tombstone_compaction_interval". If I am
> dealing with a big partition in LCS which is receiving new records every
> day,
> and a weekly incremental repair job continously anticompacting the data
> and
> thus creating SStables, what is the likelhood of the default interval
> (10 days) to be actually hit?
>
>
> It will be hit, but probably only in the repaired data. Once the data is
> marked repaired, it shouldn't be anticompacted again, and should get old
> enough to pass the compaction interval. That shouldn't be an issue though,
> because you should be running repair often enough that data is repaired
> before it can ever get past the gc grace period. Otherwise you'll have
> other problems. Also, keep in mind that tombstone eviction is a part of all
> compactions, it's just that occasionally a compaction is run specifically
> for that purpose. Finally, you probably shouldn't run incremental repair on
> data that is deleted. There is a design flaw in the incremental repair used
> in pre-4.0 of cassandra that can cause consistency issues. It can also
> cause a *lot* of over streaming, so you might want to take a look at how
> much streaming your cluster is doing with full repairs, and incremental
> repairs. It might actually be more efficient to run full repairs.
>
> Hope that helps,
>
> Blake
>
> On May 11, 2017 at 7:16:26 AM, Stefano Ortolani (ostefano@gmail.com)
> wrote:
>
> Hi all,
>
> I am trying to wrap my head around how C* evicts tombstones when using LCS.
> Based on what I understood reading the docs, if the ratio of garbage
> collectable tomstones exceeds the "tombstone_threshold", C* should start
> compacting and evicting.
>
> I am quite puzzled however by what might happen when dealing with range
> tombstones. In that case a single tombstone might actually stand for an
> arbitrary number of normal tombstones. In other words, do range tombstones
> contribute to the "tombstone_threshold"? If so, how?
>
> I am also a bit confused by the "tombstone_compaction_interval". If I am
> dealing with a big partition in LCS which is receiving new records every
> day,
> and a weekly incremental repair job continously anticompacting the data
> and
> thus creating SStables, what is the likelhood of the default interval
> (10 days) to be actually hit?
>
> Hopefully somebody will be able to shed some lights here!
>
> Thanks in advance!
> Stefano
>
>

--001a113e07d8237031054f539ac5
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Thanks a lot Blake, that definitely helps!</div><div>=
<br></div><div>I actually found a ticket re range tombstones and how they a=
re accounted for:=C2=A0<a href=3D"https://issues.apache.org/jira/browse/CAS=
SANDRA-8527" target=3D"_blank">https://issues.apache.org/<wbr>jira/browse/C=
ASSANDRA-8527</a></div><div><br></div><div>I am wondering now what happens =
when a node receives a read request. Are the range tombstones read before s=
canning the SStables? More interestingly, given that a single partition mig=
ht be split across different levels, and that some range tombstones might b=
e in L0 while all the rest of the data in L1, are all the tombstones prefet=
ched from _all_ the involved SStables before doing any table scan?</div><di=
v><br></div><div>Regards,</div><div>Stefano</div></div><div class=3D"gmail_=
extra"><br><div class=3D"gmail_quote">On Thu, May 11, 2017 at 7:58 PM, Blak=
e Eggleston <span dir=3D"ltr">&lt;<a href=3D"mailto:beggleston@apple.com" t=
arget=3D"_blank">beggleston@apple.com</a>&gt;</span> wrote:<br><blockquote =
class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid=
;padding-left:1ex"><div style=3D"word-wrap:break-word"><div id=3D"m_-512460=
5967271343458bloop_customfont" style=3D"font-family:Helvetica,Arial;font-si=
ze:12px;color:rgba(0,0,0,1.0);margin:0px;line-height:auto">Hi Stefano,</div=
><span class=3D""><div id=3D"m_-5124605967271343458bloop_customfont" style=
=3D"font-family:Helvetica,Arial;font-size:12px;color:rgba(0,0,0,1.0);margin=
:0px;line-height:auto"><br><blockquote type=3D"cite" class=3D"m_-5124605967=
271343458clean_bq"><div dir=3D"ltr"><div>Based on what I understood reading=
 the docs, if the ratio of garbage=C2=A0</div><div>collectable tomstones ex=
ceeds the &quot;tombstone_threshold&quot;, C* should start=C2=A0</div><div>=
compacting and evicting.</div></div></blockquote></div> <div><br></div></sp=
an><div><div id=3D"m_-5124605967271343458bloop_sign_1494526774315814144" cl=
ass=3D"m_-5124605967271343458bloop_sign"></div> If there are no other norma=
l compaction tasks to be run, LCS will attempt to compact the sstables it e=
stimates it will be able to drop the most tombstones from. It does this by =
estimating the number of tombstones an sstable has that have passed the gc =
grace period. Whether or not a tombstone will actually be evicted is more c=
omplicated. Even if a tombstone has passed gc grace, it can&#39;t be droppe=
d if the data it&#39;s deleting still exists in another sstable, otherwise =
the data would appear to return. So, a tombstone won&#39;t be dropped if th=
ere is data for the same partition in other sstables that is older than the=
 tombstone being evaluated for eviction.<span class=3D""><div><br></div><di=
v><blockquote type=3D"cite" class=3D"m_-5124605967271343458clean_bq"><div d=
ir=3D"ltr"><div>I am quite puzzled however by what might happen when dealin=
g with range=C2=A0</div><div>tombstones. In that case a single tombstone mi=
ght actually stand for an=C2=A0</div><div>arbitrary number of normal tombst=
ones. In other words, do range tombstones=C2=A0</div><div>contribute to the=
 &quot;tombstone_threshold&quot;? If so, how?</div></div></blockquote></div=
><div><br></div></span><div>From what I can tell, each end of the range tom=
bstone is counted as a single tombstone tombstone. So a range tombstone eff=
ectively contributes &#39;2&#39; to the count of tombstones for an sstable.=
 I&#39;m not 100% sure, but I haven&#39;t seen any sstable writing logic th=
at tracks open tombstones and counts covered cells as tombstones. So, it=
9;s likely that the effect of range tombstones covering many rows are under=
 represented in the droppable tombstone estimate.</div><span class=3D""><di=
v><br></div><div><blockquote type=3D"cite" class=3D"m_-5124605967271343458c=
lean_bq"><div dir=3D"ltr"><div>I am also a bit confused by the &quot;tombst=
one_compaction_interval<wbr>&quot;. If I am=C2=A0</div><div>dealing with a =
big partition in LCS which is receiving new records every day,=C2=A0</div><=
div>and a weekly incremental repair job continously anticompacting the data=
 and=C2=A0</div><div>thus creating SStables, what is the likelhood of the d=
efault interval=C2=A0</div><div>(10 days) to be actually hit?</div></div></=
blockquote></div><div><br></div></span><div>It will be hit, but probably on=
ly in the repaired data. Once the data is marked repaired, it shouldn&#39;t=
 be anticompacted again, and should get old enough to pass the compaction i=
nterval. That shouldn&#39;t be an issue though, because you should be runni=
ng repair often enough that data is repaired before it can ever get past th=
e gc grace period. Otherwise you&#39;ll have other problems. Also, keep in =
mind that tombstone eviction is a part of all compactions, it&#39;s just th=
at occasionally a compaction is run specifically for that purpose. Finally,=
 you probably shouldn&#39;t run incremental repair on data that is deleted.=
 There is a design flaw in the incremental repair used in pre-4.0 of cassan=
dra that can cause consistency issues. It can also cause a *lot* of over st=
reaming, so you might want to take a look at how much streaming your cluste=
r is doing with full repairs, and incremental repairs. It might actually be=
 more efficient to run full repairs.</div><div><br></div><div>Hope that hel=
ps,</div><div><br></div><div>Blake</div><div><div class=3D"h5"><div><br><di=
v><div><p class=3D"m_-5124605967271343458airmail_on">On May 11, 2017 at 7:1=
6:26 AM, Stefano Ortolani (<a href=3D"mailto:ostefano@gmail.com" target=3D"=
_blank">ostefano@gmail.com</a>) wrote:</p> <blockquote type=3D"cite" class=
=3D"m_-5124605967271343458clean_bq"><span><div><div></div><div>


<div dir=3D"ltr">
<div>Hi all,</div>
<div><br></div>
<div>I am trying to wrap my head around how C* evicts tombstones
when using LCS.</div>
<div>Based on what I understood reading the docs, if the ratio of
garbage=C2=A0</div>
<div>collectable tomstones exceeds the &quot;tombstone_threshold&quot;, C*
should start=C2=A0</div>
<div>compacting and evicting.</div>
<div><br></div>
<div>I am quite puzzled however by what might happen when dealing
with range=C2=A0</div>
<div>tombstones. In that case a single tombstone might actually
stand for an=C2=A0</div>
<div>arbitrary number of normal tombstones. In other words, do
range tombstones=C2=A0</div>
<div>contribute to the &quot;tombstone_threshold&quot;? If so, how?</div>
<div><br></div>
<div>I am also a bit confused by the
&quot;tombstone_compaction_interval<wbr>&quot;. If I am=C2=A0</div>
<div>dealing with a big partition in LCS which is receiving new
records every day,=C2=A0</div>
<div>and a weekly incremental repair job continously anticompacting
the data and=C2=A0</div>
<div>thus creating SStables, what is the likelhood of the default
interval=C2=A0</div>
<div>(10 days) to be actually hit?</div>
<div><br></div>
<div>Hopefully somebody will be able to shed some lights
here!</div>
<div><br></div>
<div>Thanks in advance!=C2=A0</div>
<div>Stefano=C2=A0</div>
<div><br></div>
</div>


</div></div></span></blockquote></div></div></div></div></div></div></div><=
/blockquote></div><br></div>

--001a113e07d8237031054f539ac5--