Manipulating contents of a file for determining deadlock - bash

Question

I've been working on a short assignment to determine deadlocks in database soon after it is identified. For this i will be regularly monitoring the process details once i dump the same to a text file.
After certain manipulation The output file (a miniature version) to be monitored is:
INPUT FILE:-
SPID BLOCKING SPID
23 50
71 46
50 60
60 96
This means that process 23 is being blocked by process 50 and in turn by 60.
So in this case the process chain dependency is 23->50->60->96 and the other 71->46
In short I want the final culprit spids.. in this case 96 and 46 along with the chain count ( 4 and 2)
OUTPUT FILE:-
CULPRIT SPID CHAIN_COUNT
1) 96 4
2) 46 2
For this the entire file will have to be looped and compare column 2 matching with any column 1 in the file,. This has to be repeated over until there is no match for column 2
I need to acheive this via awk or sed, however I believe this can be achieved via linked
list.
Any kind of suggesstion are welcom
****JS just one more question, for sample file
SPID BLOCKING SPID
45 11
12 34
34 35
23 60
71 45
60 71
OUTPUT I RECEIVE:
71=>45
23=>60->71
12=>34->35
OUTPUT I AM SUPPOSED TO GET
23=>60=>71=>45
12=>34->35
I hope you got what I am trying to say.

Answer 1

Let's try to solve this in two steps. Step one is where we shall create the chain using your input file (assuming that it might be useful for you to go back and check the trail) :
awk '
NR>1 {
for(value in values) {
if ($1 == values[value]) {
chain[$1] = chain[value] "->" $2;
values[$1] = $2;
delete chain[value];
delete values[value];
next
}
}
values[$1] = $2;
chain[$1] = $1 "->" $2
}
END {
for(left in values) {
for(c in chain) {
x = split(chain[c],t,/->/)
if (left == t[x]) {
chain[left] = chain[c] "->" values[left]
delete chain[c]
values[left]
}
}
}
for (trail in chain) print chain[trail]
}' file
For your input file, this gives the output as:
23->60->71->45->11
12->34->35
Now we will pipe our above script to another awk for formatting.
awk -F'->' '
BEGIN {
OFS="\t";
print "CULPRIT SPID", "CHAIN_COUNT"
}
{
print $NF, NF
}'
This will give the output as:
CULPRIT SPID CHAIN_COUNT
11 5
35 3
As I said before assuming you need to look at the trail. If trail is of no use to you then you can simply do the following:
awk '
BEGIN {
OFS="\t";
print "CULPRIT SPID", "CHAIN_COUNT"
}
NR>1 {
for(value in values) {
if ($1 == values[value]) {
chain[$1] = chain[value] FS $2;
values[$1] = $2;
delete chain[value];
next
}
}
values[$1] = $2;
chain[$1] = $1 FS $2
}
END {
for(left in values) {
for(c in chain) {
x = split (chain[c],t)
if (left == t[x]) {
chain[left] = chain[c] FS values[left]
delete chain[c]
values[left]
}
}
}
for(v in chain) {
num = split(chain[v], culprit)
print culprit[num], num
}
}' file
Output will be:
CULPRIT SPID CHAIN_COUNT
11 5
35 3
You can re-direct it to an output file as you please.