Category Archives: Powershell

Running a plethora of T-SQL

Have you ever needed to run a bunch of scripts, over and over and over? And the SSMS tabs – it can be a mess. This is mind-numbing work – call in the computers. Check out this script I forked from GitHub and modified: https://github.com/propellor/Powershell-pasen/blob/master/Run-Sql.ps1

You point the script it at a folder of scripts and it will fire them off using invoke-sqlcmd. You can use it in a sort-of debugging mode, to get those scripts all working when you get a mess from a co-worker or vendor and have it pick up rerunning from the failed script. Alternatively, if someone else is responsible for the fixin’ and you are just the runner, you can have all the failed scripts rename themselves to .failure.

Also, my first GitHub fork.

UPDATE: Since doing this, I’ve started using RoundhousE, which does the same thing, but waaaay better. Using TFS and octopus deploy, we package up roundhouse with some sql scripts and let developers deploy database scripts to staging and then production.

Creating Restore Statements With Powershell

WARNING: Use caution in running scripts you find on the internet. You are responsible for the well being of your systems!

We’re sitting around waiting for the SAN monkeys to fling some crap our way – not the SAN guys, the SAN monkeys, kinda like gremlins, but not as bad. Our primary cluster is down, which runs 100+ databases. Maybe the disks will come up – or maybe they won’t.

Assessing the situation, I found that the MSDB was not in full recovery mode. People, please don’t run around changing DBs to simple to ‘save space’ without thinking the ramifications through. In this case, we lose all backup history so we now need to figure out what backup file is the most recent and then which log files come after that .bak file – then we slam them together into a file, leaving the NORECOVERY clause off the last log restore statement. If you think I’m doing this by hand, you’re crazy.

We make an assumption here. Each file holds only one backup set. This is how we do our backup files, so I’m OK with this assumption.


$DatabaseBackupRoot = “\\MYSERVA\e$\Backups\"
$fix = ""

$DatabaseBackupDirs = gci $DatabaseBackupRoot

foreach ($DatabaseBackupDir in $DatabaseBackupDirs)
{
$DB = $DatabaseBackupDir.Name

$BackupRestore = ""
$LogRestore = ""
$LogRestore = ""
$BackupRestore = ""
# Parameters

$Path = “\\MYSERVA\e$\Backups\$DB”;
$FullPattern = “*.bak”;
$LogPattern = “*.trn”;

# Enumerate Last Full Back
$FullBackup = Get-ChildItem $Path -Filter ($DB + $FullPattern) `
| sort Name `
| Select-Object -Last 1;

# Enumerate Log Backups
$LogBackups = Get-ChildItem $Path -Filter ($DB + $LogPattern) `
| Where-Object { $_.LastWriteTime -gt $Fullbackup.LastWriteTime} `
| sort Name;

$LogBackupsCount = $LogBackups.Count

$BackupFileName = $FullBackup.Name
# if the backupfile is null, skip this
if ($BackupFileName)
{

if ($LogBackupsCount > 0)
{
$BackupRestore = "RESTORE DATABASE [$db] FROM DISK = N'$DatabaseBackupRoot$DB\$BackupFileName' WITH FILE = 1, NORECOVERY, NOUNLOAD, STATS = 10
GO
"
}
else
{
$BackupRestore = "RESTORE DATABASE [$db] FROM DISK = N'$DatabaseBackupRoot$DB\$BackupFileName' WITH FILE = 1, NOUNLOAD, STATS = 10
GO
"
}

}
else{
#We get a list in the output of emtpy directories.
Write-Output $DB
}

$counter = 0
$len = $LogBackups.Count

foreach ($LogBackup in $LogBackups)
{
$counter++

$LogBackupFile = $LogBackup.Name;
#Test for no log/empty directory
if($LogBackupFile)
{
if($counter -lt $len)
{
$LogRestore = $LogRestore + "
RESTORE LOG [$db] FROM DISK = N'$DatabaseBackupRoot$DB\$LogBackupFile' WITH FILE = 1, NORECOVERY, NOUNLOAD, STATS = 10
GO
"
}
else
{
$LogRestore = $LogRestore + "
RESTORE LOG [$db] FROM DISK = N'$DatabaseBackupRoot$DB\$LogBackupFile' WITH FILE = 1, NOUNLOAD, STATS = 10
GO
"

}
}

}

$RestoreOneDatabase = $BackupRestore + $LogRestore

$RestoreAllDatabases = $RestoreAllDatabases + $RestoreOneDatabase

}

$RestoreAllDatabases | out-file "c:\temp\restore-MYSERVA.sql"

So now we’ve got a SQL statement with all these restores compiled together and separated with GO statements. Pretty neat. I’d probably run small blocks of them and fix any issues the script might have.

Monitoring SSAS for process failures

When a cube fails to process, you don’t want to hear it from your customer. As the last step in our datawarehouse load, I was processing a cube, using the script generated from the SSMS dialog. That was working fine most of the time, but sometimes we’d have a development failure after some changes had been made during the day.

Here is what worked. Notes on this – don’t use $( anywhere in your PS code or SQL will parse it as a token and tell you to escape it. My first thought was to send an email to DBAs when the process failed – or rather hadn’t happened within the last 12 hours, but opted for the throw instead. Throw raises a nice error in the job and brings it to the attention of our central monitoring system.


## Add the AMO namespace
$loadInfo = [Reflection.Assembly]::LoadWithPartialName(“Microsoft.AnalysisServices”)

## Connect and get the edition of the local server
$connection = “SERVER01”
$server = New-Object Microsoft.AnalysisServices.Server
$server.connect($connection)

$databases = $server.get_Databases()
$WF_DB = $databases.GetByName("SSASDB1")

$d = Get-Date #did this because SQL job step doesn't like dollar sign(get-date).AddHours...
$date = $d.AddHours(-12)

if ( $WF_DB.LastProcessed -lt $date)
{# Cube not updated since our threshold date

$dateprocessed = $WF_DB.LastProcessed
$lastupdate = $WF_DB.LastUpdate
$parent = $WF_DB.Parent

$message = "SSASDB1 on $parent has not processed since $date. The last process was on $dateprocessed. Manually reprocess the cube to see error message or view the previous job step for additonal information. Thanks!"
throw $message
}
else
{
echo "Cube updated after $date"

}

SQL Powershell

I’m getting a little tired of always clicking around in SSMS. Here goes a shot at powershell. I’ve done a few tutorials on powershell, but don’t use it enough to remember. The only thing I remember is Get-Child-Items and it’s alias GCI. That’s a start. Follow along with me.

User asks if a database was backed up last night. Of course it was, but let’s just make sure.

Right click on the databases node of the server I’m interested about in ssms and ‘Start PowerShell’

I end up at the databases path and try to copy the text out of the window with a mouse swipe to paste it in here. No dice, so I pop into properties and enable quick edit mode – I prefer it on in the CMD window. A highlight and a right-click will copy text out of the window.

PS SQLSERVER:\SQL\SERVER1\DEV08\Databases>

You can see the Server, instance name and then the databases node.

I enter my one command: GCI

A long list of databases fill the screen, but not the backup information I’m looking for. I know there is a powershell cmdlet (command-let a powershell command object) which will tell me more about the children of the database node – the databases. Fortunately, everyone on the web likes to talk about it. It’s the Get-Member cmdlet, which will tell you the methods and properties of whatever object you feed or ‘pipe’ it. Let’s pipe those database objects to the Get-Member alias.

GCI | GM

Whoah. That’s alot of output. I could use a little help sorting through that. Maybe I can pipe that list to another cmdlet and search for a string. There is probably a cmdlet called Find.

GCI | GM | FIND ‘backup’
FIND: Parameter format not correct

So how do I know what a cmdlet is expecting?

find /?

That gives me some help info, but changing the single quotes to double quotes doesn’t seem to do the trick.

Get-Help find

There’s a listing of somewhat unrelated commands. What did I just do there? What happens if I don’t give it a parameter.

Get-Help

TOPIC
Get-Help

SHORT DESCRIPTION
Displays help about Windows PowerShell cmdlets and concepts.

LONG DESCRIPTION

SYNTAX
get-help { | }
help { | }
-?

I expected to see a help page a bit more like this when I typed in Get-Help find, but instead I saw a listing. Find is being interpreted as a TopicName. So let’s type Get-Help find again – maybe there’s a more appropriate cmdlet I can use to find my search string in output of Get-Member. Select-String stands out to me – Finds text in strings and files.

GCI| GM | Select-String “Backup”

System.Void DropBackupHistory()
System.Data.DataTable EnumBackupSetFiles(int backupSetID), System.Data.DataTable EnumBackupSetFiles()
System.Data.DataTable EnumBackupSets()
System.Boolean IsDbBackupOperator {get;}
System.DateTime LastBackupDate {get;}
System.DateTime LastDifferentialBackupDate {get;}
System.DateTime LastLogBackupDate {get;}

There it is – LastBackupDate – and even more.

So I’ll accomplish my task at hand.

GCI | SELECT NAME, LASTBACKUPDATE, LASTLOGBACKUPDATE

But I’m only interested in a couple databases which start with “Risk”.

GCI | where { $_.name -like “Risk*”} | SELECT NAME, LASTBACKUPDATE, LASTLOGBACKUPDATE

The $_ represents the object being sent from the previous cmdlet. So you can filter on any property you found in the get-members listing.

Mission accomplished! That gives me what I need and I can feel smug about not clicking around in SSMS. True, it took me five minutes to learn how to do this and would take me a little longer than clicking on the two databases which were returned, but learning this will let me complete other more complex tasks.

Oh yeah, one other thing. Out-Host -Page will let you page through the sometimes lengthy manual pages. OH -p also works.