Unable to connect to Redshift on macOS Sierra about odbc

[jimhester]

Perhaps try UTF-16? Clearly UTF-8 is not correct...

[jimhester]

The single letter table names indicate that a mismatch between the encoding returned by the driver and what unixODBC is expecting.

In the amazon.redshiftodbc.ini file, located by default in /opt/amazon/redshift/lib/amazon.redshiftodbc.ini, you need to use the following settings

DriverManagerEncoding=UTF-8
ODBCInstLib=libodbcinst.so

You also may want to set a valid logPath

[jimhester]

The amazon instructions you mention seem to be using iODBC, not unixODBC, perhaps that is part of the issue. The configuration instructions also mention modifying your LD_LIBRARY_PATH to ensure the driver libraries are loaded properly. You should ensure this is the case, particularly if you are starting R from somewhere other than the command line (like the RStudio IDE).

Also I would suggest using exactly the same strings in the manual configuration as in your odbc.ini, the default strings will likely not work for redshift.

A successful connection with isql should mean that odbc should also work, it is difficult to diagnose the issue without having a redshift instance to test against.

[kylelundstedt]

Thanks for the quick reply, @jimhester. I made progress by substituting in the actual path to the Amazon Redshift driver for the driver parameter. The good news is that the connection succeeds.

However, I get bogus answers when I use the odbc functions. I do have multiple schemas in the mortgage database; per issue #18, is there a workaround to use schema names in the odbc functions?

> library(DBI); library(ggplot2)
> con <- dbConnect(
+   odbc::odbc(),
+   driver = "/opt/amazon/redshift/lib/libamazonredshiftodbc.dylib",
+   servername = "mortgage.shared-vault.com", database = "mortgage",
+   UserName = "xxx", Password = "yyy", port = 5439
+ )
> dbListTables(con)
 [1] "c" "c" "d" "d" "g" "g" "l" "l" "l" "p" "a" "a" "c" "d" "e"
[16] "e" "h" "h" "l" "l" "l" "o" "p" "p" "p" "r" "s" "c" "f" "c"
[31] "i" "l" "l" "q" "s" "h" "h" "f" "p" "s" "s" "t" "l" "l" "s"
[46] "v" "i" "i" "i" "i" "i" "m" "m" "m" "m" "m" "z" "c" "l" "c"
[61] "c" "d" "d" "d" "g" "l" "l" "l" "l" "a" "f" "h" "h" "h" "h"
[76] "h" "h" "p" "p"
> dbListTables(con, table_name = "f%")
[1] "f" "f" "f"
> mpg_table <- dbWriteTable(con, "mpg", mpg)
Error in new_result(connection@ptr, statement) : 
  nanodbc.cpp:1587: 42501: [Amazon][Amazon Redshift] (30) Error occurred while trying to execute a query: [SQLState 42501] ERROR:  permission denied for schema cfpb_hmda

[kylelundstedt]

Hmm. I edited /opt/amazon/redshift/lib/amazon.redshiftodbc.ini.

[Driver]
## - Note that this default DriverManagerEncoding of UTF-32 is for iODBC.
# DriverManagerEncoding=UTF-32
DriverManagerEncoding=UTF-8
ErrorMessagesPath=/opt/amazon/redshift/ErrorMessages
LogPath=[LogPath]
SwapFilePath=/tmp

#   iODBC
# ODBCInstLib=libiodbcinst.dylib

ODBCInstLib=libodbcinst.so

I then got this output. 😄 Any ideas?

> dbListTables(con)
 [1] "畣楳彰楬歮"                    
 [2] "畣楳彰楬歮慟汬损獵灩"        
...

[kylelundstedt]

UTF-16 worked like a charm; awesome!!!

If I need to read or write to a specific schema, is that possible?

[jimhester]

I need to look into schema support a bit more I think, but you can always fall back to writing the SQL by hand if needed and passing it to dbGetQuery().

[kylelundstedt]

Sounds good; thanks for all your help!

[ftoresh]

I have the same issue with Athena. Any idea?

I have the same issue with Athena. Any idea?
Perhaps try UTF-16 (as jimhester has recommended)

[ftoresh]

I have the same issue with Athena. Any idea?
Perhaps try UTF-16 (as jimhester has recommended)

Indeed it worked!