XML Log file regex

Question

A legacy system I cannot change is pumping out 5 Gig of mostly awful XML logs per day and blowing my ingestion licence. There are 2 classes of verbose errors occurring 1000+ times a minute, but every few minutes one genuinely interesting entry. I'd like to drastically shorten the repeating entries in sed, and retain the interesting ones untouched

So what I need
1. Regexes to match each of the 2 classes of annoying log entry (eg ...'decimal'... and ...'DBNull'... but not the occasional interesting ones).
One regex to match each annoying error class is fine, I can do 2 sed passes
2. I need a capture group with the timestamp so I can replace the long XML lines with a succinct version - but with the correct timestamp so as not to lose fidelity.

I've gotten as far as this to match the and the capture created date:

(?:<Log).*?(createdDate="\d{2}\/\d{2}\/\d{4}.\d{2}:\d{2}:\d{2}").*?(?:decimal).*?(<\/Log>)

which is close, but suffers from a kind of reverse greediness where I match the from 'decimal' to a an opening Log statement several entries earlier Have played around negative look-behind but just given myself a severe headache

Sample Data

<Log type="ERROR" createdDate="11/09/2015 08:13:14" > 
 <![CDATA[ [108] -- much cruft removed-- SerializationException: There was an error deserializing the object of type Common.DataCtract.QResult. The value '' cannot be parsed as the type 'decimal'. ---> System.Xml.XmlException: The value '' cannot be parsed as the type 'decimal'. ---> System.FormatException: Input string was not in a correct format.
  ]]></Log> 

<Log type="ERROR" createdDate="11/09/2015 08:13:13" > 
 <![CDATA[ [108] -- much cruft removed-- SerializationException: There was an error deserializing the object of type Common.DataCtract.QResult. The value '' cannot be parsed as the type 'decimal'. ---> System.Xml.XmlException: The value '' cannot be parsed as the type 'decimal'. ---> System.FormatException: Input string was not in a correct format.
  ]]></Log> 

<Log type="ERROR" createdDate="11/09/2015 08:13:12" > 
 <![CDATA[ [129] Services.DService.D.FailedToAddRQ(Exceptionex, RQEntityrQ, RHeaderEntityrHeader, StringPRef, ): FailedToAddRQ()...with parameters [pRef:=123,0,1], [rQ.AffinityCode:=],[Q.thing=thing][rQ.AffinityRQDT:=123],[rHeader.RHeaderIDPK:=123],[rQ.UWriteIDFK:=] 
  Data.DataAccessLayerException: Conversion from type 'DBNull' to type 'Long' is not valid.
Parameters:
 [RETURN_VALUE][ReturnValue] Value: [0]
 ---> System.InvalidCastException: Conversion from type 'DBNull' to type 'Long' is not valid.
 ]]></Log> 

 <Log type="ERROR" createdDate="11/09/2015 08:13:11" > 
 <![CDATA[ [129] Services.DService.D.FailedToAddRQ(Exceptionex, RQEntityrQ, RHeaderEntityrHeader, StringPRef, ): FailedToAddRQ()...with parameters [pRef:=123,0,1], [rQ.AffinityCode:=],[Q.thing=thing][rQ.AffinityRQDT:=123],[rHeader.RHeaderIDPK:=123],[rQ.UWriteIDFK:=] 
  Data.DataAccessLayerException: Conversion from type 'DBNull' to type 'Long' is not valid.
  ]]></Log> 

 <Log type="ERROR" createdDate="11/09/2015 08:13:10" > 
 <![CDATA[ [231] An actual interesting log entry with a real error message ]]></Log>

<Log type="ERROR" createdDate="11/09/2015 08:13:09" > 
 <![CDATA[ [108] -- much cruft removed-- SerializationException: There was an error deserializing the object of type Common.DataCtract.QResult. The value '' cannot be parsed as the type 'decimal'. ---> System.Xml.XmlException: The value '' cannot be parsed as the type 'decimal'. ---> System.FormatException: Input string was not in a correct format.
  ]]></Log> 

Answer 1

Not sure what you are exaclty looking for, but this is an example of how you can isolate <Log...</Log> blocks and proceed to a replacement:

sed '/^<Log /{:a;/<\/Log>/!{N;ba;};s/>.*\(decimal\|DBNull\).*</>\1</}' file.log

details:

/^<Log / { # condition: a line that starts with "<Log "
    :a;    # define the label "a"
    /<\/Log>/! { # condition: if the line doesn't contain "</Log>"
        N;       # append the next line to the pattern space
        ba;      # go to the label "a"
    };
    s/>.*\(decimal\|DBNull\).*</>\1</ # replace the block
}

(I assumed <Log are always at the start of the line unlike records at sec 10 and 11, that are probably typos.)