on 2009 Dec 16 6:35 PM
I am reliably getting this error: Failed to decompress a compressed packet when both COMP=YES and values for PBUF & PROWS are specified in a connection string.
It seems to depend on the relative values of PROWS & PBUF used, and the size of the result set. It seems that once PBUF is more than approx 100k, any value of PROWS more than about 40 will cause the error, with a decent size result set (1000 rows+). There is no problem with any combination of PBUF and PROWS if COMP=NO.
The behaviour is the same on the latest ebf of 9.0.2 and 10.0.1.3680 (server), 10.0.1.3931 (client). I can't find any refrence to a similar problem in the 10.0.1 fix list.
An example of the error from the client log:
18:15:01 [ 22] Failed to decompress a compressed packet
18:15:01 [ 22] Received bad packet (type: 24, received length: 1062, header length: 1062), disconnecting
18:15:01 [ 22] Connection terminated abnormally; error code 100001
18:15:01 [ 22] Communication function i_cs_HandleSQLPresError code 5
18:15:01 [ 22] Communication function StrmGetInd code 2
18:15:01 [ 22] Client disconnected
At the server end you just get:
Connection terminated abnormally
I've found this while testing wide area performance, for which both prefetching and compression are clearly important. I wonder if anyone else has seen something similar?
Request clarification before answering.
Fixed in 10.0.1.4116 Many thanks to everyone involved in identifying and sorting the problem.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
A good idea to report here when bugs have been fixed, obviously:)
A couple of quick tests (using comp=yes;pbuf=150k;prows=50, selecting 5000 rows of ~5000 bytes each) worked properly for me. I would recommend that you open an issue with tech support. If you give them a repro, they should be able to help diagnose the problem.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Does tech support have a mechanism for easily testing "wide area performance", which may be a critical phrase in Justin's description.
User | Count |
---|---|
87 | |
10 | |
9 | |
8 | |
6 | |
6 | |
6 | |
5 | |
5 | |
5 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.