-
zas
ask on #postgresql perhaps ;)
2016-05-24 14509, 2016
-
ruaok
zas: no. dbmirror was pre-slony, pre-streaming replication. but it makes up the backbone of our lazy replication system.
2016-05-24 14517, 2016
-
ruaok
I know C. I'm not afraid to dive in and fix what is needed.
2016-05-24 14525, 2016
-
ruaok
bitmap is on the right track.
2016-05-24 14528, 2016
-
zas
bitmap: do you know which part of pending.c writes garbarge ?
2016-05-24 14534, 2016
-
zas
garbage*
2016-05-24 14554, 2016
-
bitmap
nope, no idea yet
2016-05-24 14541, 2016
-
zas
can you compiled it with DEBUG_OUTPUT ?
2016-05-24 14557, 2016
-
zas
2016-05-24 14530, 2016
-
ruaok
2016-05-24 14549, 2016
-
ruaok
I wonder how long those lines in the output are. 256 bytes? :)
2016-05-24 14526, 2016
-
ruaok
bitmap: cron is enabled, yes?
2016-05-24 14534, 2016
-
bitmap
no, still disabled
2016-05-24 14536, 2016
-
ruaok
we've had a run (or should've) and no data in dbmirror
2016-05-24 14539, 2016
-
ruaok
ah, ok.
2016-05-24 14522, 2016
-
ruaok
go ahead and enable it.
2016-05-24 14532, 2016
-
ruaok
it is going to take a while to sort the pending stuff out.
2016-05-24 14552, 2016
-
bitmap
okay
2016-05-24 14543, 2016
-
ruaok
that replication packet data looks speshul.
2016-05-24 14505, 2016
-
ruaok
2016-05-24 14544, 2016
-
ruaok
truncated after 64 characters.
2016-05-24 14548, 2016
-
ruaok
solid clue
2016-05-24 14558, 2016
-
ruaok
sorry, bytes.
2016-05-24 14518, 2016
-
ruaok
for debugging this, I am going to create a VM and a test DB.
2016-05-24 14522, 2016
-
ruaok
going to be much more sane.
2016-05-24 14552, 2016
-
ruaok
ha. it takes longer to make a sandwich than to download ubuntu LTS on my connection. :)
2016-05-24 14556, 2016
-
bitmap
now you're compelled to make a sandwich every time you download ubuntu
2016-05-24 14501, 2016
-
ruaok
pavlov demands it!
2016-05-24 14545, 2016
-
UmkaDK_ joined the channel
2016-05-24 14555, 2016
-
UmkaDK has quit
2016-05-24 14513, 2016
-
ruaok
bitmap: can you tell if cron ran?
2016-05-24 14516, 2016
-
ruaok
no data in dbmirror
2016-05-24 14520, 2016
-
bitmap
was just doing that
2016-05-24 14534, 2016
-
bitmap
May 24 13:00:01 astro CRON[5167]: (musicbrainz) CMD (/home/musicbrainz/musicbrainz-server/admin/cron/hourly.sh)
2016-05-24 14547, 2016
-
bitmap
so I guess it ran and ended pretty abruptly
2016-05-24 14501, 2016
-
bitmap
probably 'cause everything exist on DB_READ_ONLY now
2016-05-24 14506, 2016
-
bitmap
exits
2016-05-24 14519, 2016
-
ruaok
well, lets keep our eyes peeled.
2016-05-24 14533, 2016
-
ruaok
start working on the code that will touch the rows that need to have data updated.
2016-05-24 14537, 2016
-
bitmap
I can try running that cover art script manually to see if it does anything
2016-05-24 14555, 2016
-
bitmap
without the DB_READ_ONLY check
2016-05-24 14501, 2016
-
ruaok
k
2016-05-24 14507, 2016
-
ruaok
anything to try and break things again.
2016-05-24 14538, 2016
-
bitmap
yes, this is inserting tons of stuff
2016-05-24 14553, 2016
-
bitmap
into release_meta
2016-05-24 14514, 2016
-
ruaok
time to make it respect AUTHORITAY!
2016-05-24 14527, 2016
-
ruaok
well, DB_READ_ONLY anyways.
2016-05-24 14507, 2016
-
ruaok
2016-05-24 14552, 2016
-
ruaok
do I need to do anything special to get PG 9.5 via packages on 14.04 LTS?
2016-05-24 14509, 2016
-
ruaok
zas, bitmap ^^
2016-05-24 14533, 2016
-
bitmap
2016-05-24 14542, 2016
-
bitmap
there's shell script there that can do it
2016-05-24 14521, 2016
-
ruaok
works, thanks.
2016-05-24 14545, 2016
-
ruaok
how does one install dbmirror these days? got some sql to cut and paste for me?
2016-05-24 14500, 2016
-
chirlu
See INSTALL-MASTER.md (--pending flag to InitDb).
2016-05-24 14528, 2016
-
ruaok
thanks
2016-05-24 14538, 2016
-
ruaok
I appreciate all the support from you guys. <3
2016-05-24 14515, 2016
-
bitmap
stopped cron again since it's just adding more stuff I have to clean up later
2016-05-24 14540, 2016
-
ruaok
yea. since ya found the problem makes sense
2016-05-24 14534, 2016
-
ariscop has quit
2016-05-24 14527, 2016
-
mihaitish has quit
2016-05-24 14548, 2016
-
mihaitish joined the channel
2016-05-24 14507, 2016
-
ruaok
problem confirmed
2016-05-24 14508, 2016
-
ruaok
2016-05-24 14544, 2016
-
ruaok
very interesting. the code that collects the data seems to work according to the debug statements.
2016-05-24 14538, 2016
-
bitmap
for generating new packets, I think it'll be easier and less error prone to compare the slave db on lolo against production
2016-05-24 14539, 2016
-
bitmap
maybe not much easier. likely less error prone
2016-05-24 14506, 2016
-
ruaok
you mean to test once we think we've fixed the problems?
2016-05-24 14538, 2016
-
zas
ruaok: can you point me at the buggy C code if you located it ?
2016-05-24 14541, 2016
-
bitmap
I mean for generating insert/update/delete statements that can insert the appropriate data into the dbmirror tables
2016-05-24 14502, 2016
-
ruaok
bitmap: got it.
2016-05-24 14513, 2016
-
ruaok
zas: will do. still pin pointing.
2016-05-24 14543, 2016
-
bitmap
2016-05-24 14511, 2016
-
ruaok
ewww. I'm going to have to ask you to leave my flat if you download something from sourceforge.
2016-05-24 14517, 2016
-
ruaok
just kidding. :)
2016-05-24 14521, 2016
-
zas
lol
2016-05-24 14539, 2016
-
bitmap
it's from 2002, give them a break
2016-05-24 14551, 2016
-
Freso
ruaok: You should do an IP block on your router for anything resolving to something on sourceforge.
2016-05-24 14500, 2016
-
ruaok
and hippies
2016-05-24 14541, 2016
-
ruaok
zas: the packageData() function works as expected. I verified that much. which is good.
2016-05-24 14542, 2016
-
zas
packageData() very likely the place
2016-05-24 14545, 2016
-
zas
lol
2016-05-24 14553, 2016
-
ruaok
would've been bad voodoo otherwise. :)
2016-05-24 14513, 2016
-
zas
so packageData() is ok ?
2016-05-24 14520, 2016
-
ruaok
yes.
2016-05-24 14531, 2016
-
ruaok
the datablock returned by it is correct.
2016-05-24 14538, 2016
-
zas
storeKeyData then
2016-05-24 14502, 2016
-
ruaok
or storeData
2016-05-24 14540, 2016
-
bitmap
2016-05-24 14552, 2016
-
zas
storeData() but the debug log uses storeKeyData
2016-05-24 14516, 2016
-
ruaok
k
2016-05-24 14526, 2016
-
ruaok
my debug does.
2016-05-24 14535, 2016
-
ruaok
main site may not, so we need to ensure both paths are working.
2016-05-24 14521, 2016
-
Freso
reosarevok: Ping? Still up for reviewing?
2016-05-24 14513, 2016
-
ruaok
what is PointerGetDatum ?
2016-05-24 14554, 2016
-
ruaok
About 2,930 results
2016-05-24 14556, 2016
-
ruaok
woah.
2016-05-24 14508, 2016
-
ruaok
that is effective to 404, no?
2016-05-24 14520, 2016
-
ruaok
worlds fanciest pointer cast.
2016-05-24 14548, 2016
-
zas
PointerGetDatum just casts to Datum
2016-05-24 14554, 2016
-
ruaok nods
2016-05-24 14508, 2016
-
zas
src/include/postgres.h:#define PointerGetDatum(X) ((Datum) (X))
2016-05-24 14514, 2016
-
ruaok nods
2016-05-24 14523, 2016
-
ruaok gets a sinking feeling
2016-05-24 14557, 2016
-
Slurpee joined the channel
2016-05-24 14512, 2016
-
zas
ruaok: NAMEOID is the correct type ?
2016-05-24 14516, 2016
-
bitmap
apgdiff only diffs the schema, not the data. guess I'll just write perl script to do this
2016-05-24 14549, 2016
-
ruaok
correct type of what?
2016-05-24 14500, 2016
-
zas
wait, checking smt
2016-05-24 14523, 2016
-
zas
2016-05-24 14510, 2016
-
ruaok
good question.
2016-05-24 14516, 2016
-
zas
src/include/pg_config_manual.h:#define NAMEDATALEN 64
2016-05-24 14523, 2016
-
ruaok
an incorrect query is just about the only thing that could be wrong.
2016-05-24 14525, 2016
-
ruaok
DING!
2016-05-24 14530, 2016
-
ruaok
64 bytes!
2016-05-24 14557, 2016
-
zas
2016-05-24 14557, 2016
-
UmkaDK_ has quit
2016-05-24 14558, 2016
-
ruaok
next question. what should it be instead?
2016-05-24 14528, 2016
-
zas
what is cpKeyData?
2016-05-24 14543, 2016
-
ruaok
that is the pointer that points to our data.
2016-05-24 14552, 2016
-
ruaok
that needs to be massages to be sent to execp
2016-05-24 14558, 2016
-
zas
"INSERT INTO dbmirror_pendingdata (SeqId,IsKey,Data) VALUES(currval('dbmirror_pending_seqid_seq'),$1,$2)"; <-- $2 here
2016-05-24 14512, 2016
-
ruaok nods
2016-05-24 14515, 2016
-
chirlu
The Data column is varchar.
2016-05-24 14520, 2016
-
zas
hmmm
2016-05-24 14515, 2016
-
ruaok
chirlu: I get the impression it isn't the table, but the type we're using for the insertion.
2016-05-24 14529, 2016
-
chirlu
Yes.
2016-05-24 14530, 2016
-
ruaok
the NAMEOID seems to be trucating at 64 bytes.
2016-05-24 14531, 2016
-
zas
2016-05-24 14517, 2016
-
ruaok
nice find.
2016-05-24 14535, 2016
-
zas
so a change in NAMEOID semantics, not well documented, happened in 9.2, which explains why 9.1 is ok, and 9.5 breaks
2016-05-24 14559, 2016
-
ruaok
working on the change now.
2016-05-24 14533, 2016
-
UmkaDK joined the channel
2016-05-24 14535, 2016
-
ruaok
change all instances?
2016-05-24 14541, 2016
-
zas
prolly
2016-05-24 14503, 2016
-
zas
well, ... check the matching query to be sure
2016-05-24 14521, 2016
-
zas
the OID have to match the actual column type
2016-05-24 14541, 2016
-
ruaok
BOOM
2016-05-24 14543, 2016
-
ruaok
:(
2016-05-24 14550, 2016
-
reosarevok
Freso: playing Borderlands but I guess:)
2016-05-24 14555, 2016
-
zas
hmmm what did you change ?
2016-05-24 14528, 2016
-
ruaok
all refs from nameoid to varcharoid
2016-05-24 14543, 2016
-
chirlu
That’s not enough.
2016-05-24 14545, 2016
-
ruaok
I guess it needs to be more nuances than that. :)
2016-05-24 14551, 2016
-
ruaok
nuanced
2016-05-24 14521, 2016
-
zas
read the second part of the message
2016-05-24 14540, 2016
-
zas
it shows a working version, which involves more changes i think
2016-05-24 14501, 2016
-
zas
because varchar -> memalloc, cannot work with fixed buffer
2016-05-24 14524, 2016
-
zas
but the issue is located at leasrt
2016-05-24 14535, 2016
-
chirlu
