From 4f4e30eda8421c973204b45c64654781034e0490 Mon Sep 17 00:00:00 2001
From: kanoi <gitkanoi@k1k2.com>
Date: Mon, 28 Sep 2015 21:17:56 +1000
Subject: [PATCH] sql - a couple of useful script when copying DB data for
 testing

---
 sql/rollback.sh | 195 ++++++++++++++++++++++++++++++++++++++++++++++++
 sql/tabdump.sh  |  63 ++++++++++++++++
 2 files changed, 258 insertions(+)
 create mode 100755 sql/rollback.sh
 create mode 100755 sql/tabdump.sh

diff --git a/sql/rollback.sh b/sql/rollback.sh
new file mode 100755
index 00000000..c348d7ab
--- /dev/null
+++ b/sql/rollback.sh
@@ -0,0 +1,195 @@
+#!/bin/bash
+#
+# WARNING!!!
+# 1) Don't use this
+# 2) If you do use this, only use it if your "ckpool's ckdb logfiles"
+#	(CCLs) are 100% certainly OK
+#	Data in the DB will be lost if the rollback requires reloading
+#	data that is missing from the CCLs
+# 3) If you do use this, make sure ckdb is NOT running
+# 4) The rollback is only for the tables that are generated from the CCLs
+#	If you need data from other tables, see tabdump.sql
+# 5) If you need to reload missing data that may or may not be in the CCLs
+#	then you need to convince Kano to update the ckdb -y option to handle
+#	that, with the current db, and also add code to the -y option to
+#	update the DB
+#	i.e. the ckdb -y option needs updating
+#
+# The basic use of this script is for when something goes wrong with the
+#	database and the CCLs are OK - so you can roll back to before the
+#	database problem and reload all the data
+#
+# Deleting markersummaries would only be an option if you know for sure
+#	that all data is in the CCLs
+# It would be ideal to do a markersummary table dump before a rollback
+# However, -e will be OK if the CCLs aren't missing anything reqired for
+#	the reload
+#
+# In all cases, if data is missing between the start and end of the
+#  rollback, then the reload SEQ checking will report it, unless ...
+#  by coincidence the missing data is at the end of a ckpool sequence
+#  and no gaps exist in the data before the truncation point
+# e.g. seqall goes from N..M then a new SEQ starts at 0..
+#  however if there actually was data after M, but all missing exactly up
+#  to before the SEQ restart 0 point, then ckdb couldn't know this unless
+#  the data was inconsistent
+#
+# -e will of course allow you to compare all but markersummary, before and
+#    after rolling back, by manually comparing the expired record with the
+#    new replacement unexpired record - but it deletes the markersummary data
+# -e is untidy simply because it leaves around unnecesary data, if the
+#    reload after the rollback is not missing anything and succeeds
+# -e deletes all the records that were already expired so you will lose
+#    the original DB history of those records, however a reload should
+#    regenerate a similar history
+#
+# -m is the same as -e except it doesn't touch markersummary at all
+# -m will probably cause duplicate markersummary issues for ckdb
+#
+# -r ... use at your own risk :) - but is the best option if nothing is
+#    missing from your CCLs
+#    it simply deletes all reload data back to the specified point and
+#    that data should all reappear with a successful reload
+#
+usAge()
+{
+ echo "usAge: `basename $0` [-n] -e|-r|-m workinfoid"
+ echo " Preferred option is just -e as long as you aren't missing reload data"
+ echo
+ echo " -n = use 'now()' instead of the date timestamp of now, for expiry date"
+ echo
+ echo " -e = rollback by deleteing expired data and expiring unexpired data,"
+ echo "      but deleting markersummary"
+ echo
+ echo " -r = rollback by deleting all relevant data (expired or unexpired)"
+ echo " -m = rollback by deleteing expired data and expiring unexpired data,"
+ echo "      but don't touch markersummary"
+ echo "      N.B. -m is expected to cause problems with ckdb"
+ echo
+ echo " generates the sql on stdout"
+ echo
+ echo " Read `basename $0` for an explanation"
+ exit 1
+}
+#
+idctl()
+{
+ echo "\\echo 'update idcontrol $1'
+update idcontrol
+set lastid=(select max($1) from $2),
+modifydate=$n,modifyby='$idn',
+modifycode='$mc',modifyinet='$mi'
+where idname='$1';"
+}
+#
+process()
+{
+ # so all stamps are exactly the same
+ now="`date +%s`"
+ idn="rollback_${wi}_`date -d "@$now" '+%Y%m%d%H%M%S%:::z'`"
+ mc="`basename $0`"
+ mi="127.0.0.1"
+ if [ "$usenow" ] ; then
+	n="now()"
+ else
+	n="'`date -u -d "@$now" '+%Y-%m-%d %H:%M:%S+00'`'"
+ fi
+ ex="expirydate = $n"
+ unex="expirydate > '6666-06-01'"
+ oldex="expirydate < '6666-06-01'"
+ #
+ if [ "$opt" = "-r" -o "$opt" = "-e" ] ; then
+	echo "\\echo 'delete markersummary'"
+	echo "delete from markersummary where markerid in"
+	echo " (select distinct markerid from workmarkers where workinfoidend >= $wi);"
+ fi
+ if [ "$opt" = "-r" ] ; then
+	echo "\\echo 'delete marks'"
+	echo "delete from marks where workinfoid >= $wi;"
+	echo "\\echo 'delete workmarkers'"
+	echo "delete from workmarkers where workinfoidend >= $wi;"
+ else
+	echo "\\echo 'delete/update marks'"
+	echo "delete from marks where $oldex and workinfoid >= $wi;"
+	echo "update marks set $ex where $unex and workinfoid >= $wi;"
+	echo "\\echo 'delete/update workmarkers'"
+	echo "delete from workmarkers where $oldex and workinfoidend >= $wi;"
+	echo "update workmarkers set $ex where $unex and workinfoidend >= $wi;"
+ fi
+ #
+ if [ "$opt" = "-r" ] ; then
+	echo "\\echo 'delete blocks'"
+	echo "delete from blocks where workinfoid >= $wi;"
+	echo "\\echo 'delete workinfo'"
+	echo "delete from workinfo where workinfoid >= $wi;"
+ else
+	echo "\\echo 'delete/update blocks'"
+	echo "delete from blocks where $oldex and workinfoid >= $wi;"
+	echo "update blocks set $ex where $unex and workinfoid >= $wi;"
+	echo "\\echo 'delete/update workinfo'"
+	echo "delete from workinfo where $oldex and workinfoid >= $wi;"
+	echo "update workinfo set $ex where $unex and workinfoid >= $wi;"
+ fi
+ #
+ if [ "$opt" = "-r" ] ; then
+	echo "\\echo 'delete miningpayouts'"
+	echo "delete from miningpayouts where payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "\\echo 'delete payments'"
+	echo "delete from payments where payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "\\echo 'delete payouts'"
+	echo "delete from payouts where workinfoidend >= $wi;"
+ else
+	echo "\\echo 'delete/update miningpayouts'"
+	echo "delete from miningpayouts where $oldex and payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "update miningpayouts set $ex where $unex and payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "\\echo 'delete/update payments'"
+	echo "delete from payments where $oldex and payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "update payments set $ex where $unex and payoutid in"
+	echo " (select distinct payoutid from payouts where workinfoidend >= $wi);"
+	echo "\\echo 'delete/update payouts'"
+	echo "delete from payouts where $oldex and workinfoidend >= $wi;"
+	echo "update payouts set $ex where workinfoidend >= $wi and $unex;"
+ fi
+ #
+ idctl markerid workmarkers
+ idctl paymentid payments
+ idctl payoutid payouts
+ # this makes sure the worker data is consistent - don't remove this
+ idctl workerid workers
+}
+#
+if [ -z "$1" ] ; then
+ usAge
+fi
+#
+if [ "$1" = "-?" -o "$1" = "-h" -o "$1" = "-help" -o "$1" = "--help" ] ; then
+ usAge
+fi
+#
+usenow=""
+if [ "$1" = "-n" ] ; then
+ usenow="y"
+ shift
+fi
+#
+if [ "$1" != "-r" -a "$1" != "-e" -a "$1" != "-m" ] ; then
+ echo "ERR: Unknown p1='$1'"
+ usAge
+fi
+#
+opt="$1"
+shift
+#
+if [ -z "$1" ] ; then
+ echo "ERR: missing workinfoid"
+ usAge
+fi
+#
+wi="$1"
+#
+process
diff --git a/sql/tabdump.sh b/sql/tabdump.sh
new file mode 100755
index 00000000..2c684740
--- /dev/null
+++ b/sql/tabdump.sh
@@ -0,0 +1,63 @@
+#!/bin/bash
+#
+# idcontrol is updated after the rows are loaded
+#  for users, paymentaddresses and workers
+#
+# note that this doesn't save idcontrol since it
+#  would cause a consistency problem with the reload
+# Instead it generates it based on the current DB
+#
+# Obviously ... don't run the output file when CKDB is running
+#
+# initid.sh would also be useful to create any missing idcontrol
+#  records, before using this to update them
+#
+t0="optioncontrol paymentaddresses useratts users workers version"
+#
+usAge()
+{
+ echo "usAge: `basename $0` -r > log.sql"
+ echo " -r = do it"
+ echo " dump tables not part of the reload"
+ exit 1
+}
+#
+idctl()
+{
+ echo "
+update idcontrol
+set lastid=(select max($1) from $2),
+modifydate=now(),modifyby='$idn',
+modifycode='$mc',modifyinet='$mi'
+where idname='$1';"
+}
+#
+process()
+{
+ t=""
+ for i in $t0 ; do
+	t="$t -t $i"
+	echo "delete from $i;"
+ done
+ pg_dump -a $t ckdb
+ #
+ idn="tabdump-`date "+%Y%m%d%H%M%S"`"
+ mc="`basename $0`"
+ mi="127.0.0.1"
+ #
+ idctl userid users
+ idctl workerid workers
+ idctl paymentaddressid paymentaddresses
+
+ # these below are just to make sure the DB is consistent
+ idctl markerid workmarkers
+ idctl paymentid payments
+ idctl payoutid payouts
+}
+#
+if [ "$1" != "-r" ] ; then
+ echo "Missing -r"
+ usAge
+fi
+#
+process