We are also experiencing a high cpu peaks on our production server.
After researching I've found out that we have some static dictionaries in our solution..
One of them looks like this:
public static Dictionary<string, List<ID>> OrderNrLookupTable
{
get
{
const string cacheKey = "ProductRepository_OrderNrHashTable1";
var results = CacheManager.GetFromCache<Dictionary<string, List<ID>>>(cacheKey);
if (results == null)
{
results = new Dictionary<string, List<ID>>();
foreach (var orderNrItem in GetAllOrderNrs())
{
var product = orderNrItem.InnerItem.Parent;
if (product != null && product.IsProduct())
{
if (!results.ContainsKey(orderNrItem.OrderNumber))
{
results.Add(orderNrItem.OrderNumber, new List<ID> {product.ID});
}
else
{
results[orderNrItem.OrderNumber].Add(product.ID);
}
}
}
CacheManager.SaveToCache(results, cacheKey);
}
return results;
}
}
My question is; Is this thread safe? Could this cause the high cpu peaks on our production servers?
Update
The threads are not modifying the dictionary.
Related
From this question, I understand that SplitStream in Apache Flink is now deprecated and it's recommended to use side-outputs instead.
Can someone give an example of how side-output can replace the splitStream?
e.g. How can the code snippet below be modified to use side-output?
DataStream mainDataStream = some definition
SplitStream<some-type> splitStream = mainDataStream.select("some-string")
Rather than
SplitStream<Integer> split = someDataStream.split(new OutputSelector<Integer>() {
#Override
public Iterable<String> select(Integer value) {
List<String> output = new ArrayList<String>();
if (value % 2 == 0) {
output.add("even");
}
else {
output.add("odd");
}
return output;
}
});
DataStream<Integer> evens = split.select("even");
DataStream<Integer> odds = split.select("odd");
you can now do
final OutputTag<Integer> evenTag = new OutputTag<String>("even"){};
final OutputTag<Integer> oddTag = new OutputTag<String>("odd"){};
SingleOutputStreamOperator<Integer> mainDataStream = someDataStream
.process(new ProcessFunction<Integer, Integer>() {
#Override
public void processElement(
Integer value,
Context ctx,
Collector<Integer> out) throws Exception {
if (value % 2 == 0) {
ctx.output(evenTag, value);
} else {
ctx.output(oddTag, value);
}
});
DataStream<Integer> evens = mainDataStream.getSideOutput(evenTag);
DataStream<Integer> odds = mainDataStream.getSideOutput(oddTag);
Note that, unlike split streams, the side output streams can be of different types.
I have 3 Oracle databases; production, test, development. For the most part, they are all identical. In my application, I would like the changes to be applied to multiple databases. For example:
using (var context = new Context())
{
context.People.Add(new Person { name = "sean" });
context.SaveChanges();
}
I then tried to override the SaveChanges method and save to multiple databases by doing this:
public void SaveChanges(int auditPersonNumber)
{
OracleCredentials.Default.Server = "VDev";
base.SaveChanges();
OracleCredentials.Default.Server = "VTest";
base.SaveChanges();
OracleCredentials.Default.Server = "VProd";
base.SaveChanges();
}
This didn't work but should explain what I am trying to achieve.
I haven't yet used EntityFramework against an Oracle database, but it should be similar to connecting against SQL Server in that the database name is specified via a ConnectionString. Your project should have a config file (web.config, app.config, or if it's a .NET Core application it could be in appsettings.json) with that ConnectionString in it.
For example:
<add name="YourConnectionString" providerName="YourOracleProviderName" connectionString="User Id=test;Password=testpassword;Data Source=eftest" />
The DbContext base constructor accepts a string argument that specifies which ConnectionString it should use, and thus which database to connect to. If you look into your context class, the default constructor should call the base constructor with that argument.
public YourDbContext() : base("YourConnectionString") {}
In order to save to multiple databases you will need to work against different instances of DbContext each with a different ConnectionString argument. So, your config will need to list a few different connection strings for every Db and you'll probably want your DbContext class to allow the argument in its constructor as well.
Perhaps the SaveChanges method implementation could instantiate the other DbContexts you'd need to use:
public void SaveChanges(int auditPersonNumber)
{
using (var context = new Context("OtherConnectionString1"))
{
// apply same changes
context.SaveChanges();
}
using (var context = new Context("OtherConnectionString2"))
{
// apply same changes
context.SaveChanges();
}
base.SaveChanges();
}
As for the applying the same changes, I would expect you can read them out from the DbContext ChangeTracker. There's an explanation about that using EF Core here but in earlier versions it's similar: http://www.entityframeworktutorial.net/efcore/changetracker-in-ef-core.aspx
Also keep in mind that the SaveChanges call to OtherConnectionString1 could succeed while others could fail, so the data might be inconsistent in your different databases. You may have to look into using transactions across multiple databases but I haven't done this yet myself.
I was able to figure out a solution thanks to the help of Sangman.
public class Context : Shared.Data.Context
{
new public void SaveChanges(int auditPersonNumber)
{
var errors = string.Empty;
var testConnectionString = "ConnectionString";
var developmentConnectionString = "ConnectionString";
//Save to test database
if (SecurityMaintenanceUser.ApplyToTest)
errors = ApplyToDatabase(testConnectionString, auditPersonNumber, "Test");
if (!string.IsNullOrWhiteSpace(errors))
errors += "\n\n";
//Save to development database
if (SecurityMaintenanceUser.ApplyToDevelopment)
errors += ApplyToDatabase(developmentConnectionString, auditPersonNumber, "Development");
if (!string.IsNullOrWhiteSpace(errors))
MessageBox.Show(errors, "Error", MessageBoxButtons.OK, MessageBoxIcon.Error);
//Save to production database
base.SaveChanges(auditPersonNumber);
}
private string ApplyToDatabase(string connectionString, int auditPersonNumber, string server)
{
try
{
using (var context = new Context(connectionString))
{
context.Configuration.ValidateOnSaveEnabled = false;
foreach (var entry in ChangeTracker.Entries())
{
var dataSet = context.Set(entry.Entity.GetType());
if (entry.State == EntityState.Added)
{
dataSet.Add(entry.Entity);
}
else if (entry.State == EntityState.Deleted)
{
var contextEntity = dataSet.Find(GetPrimaryKeyValues(entry));
context.DeleteEntity(contextEntity, auditPersonNumber);
}
else if (entry.State == EntityState.Modified)
{
var contextEntity = dataSet.Find(GetPrimaryKeyValues(entry));
context.Entry(CopyProperties(entry.Entity, contextEntity)).State = EntityState.Modified;
}
}
context.SaveChanges(auditPersonNumber);
return string.Empty;
}
}
catch (Exception e)
{
return $"Failed to apply database changes to {server}.\n{e.GetFullMessage()}";
}
}
private object CopyProperties(object source, object destination)
{
if (source == null || destination == null)
throw new Exception("Source or/and Destination Objects are null");
var typeDest = destination.GetType();
var typeSrc = source.GetType();
foreach (var srcProp in typeSrc.GetProperties())
{
if (srcProp.Name == "Type" || srcProp.Name == "AuthenticationLog")
continue;
//This blocks any complex objects attached to the entity, will need to be changed for your application
if (srcProp.PropertyType.FullName.Contains("Library.Shared"))
continue;
if (!srcProp.CanRead)
continue;
var targetProperty = typeDest.GetProperty(srcProp.Name);
if (targetProperty == null)
continue;
if (!targetProperty.CanWrite)
continue;
if (targetProperty.GetSetMethod(true)?.IsPrivate == true)
continue;
if ((targetProperty.GetSetMethod().Attributes & MethodAttributes.Static) != 0)
continue;
if (!targetProperty.PropertyType.IsAssignableFrom(srcProp.PropertyType))
continue;
targetProperty.SetValue(destination, srcProp.GetValue(source, null), null);
}
return destination;
}
private object GetPrimaryKeyValues(DbEntityEntry entry)
{
var objectStateEntry = ((IObjectContextAdapter)this).ObjectContext.ObjectStateManager.GetObjectStateEntry(entry.Entity);
return objectStateEntry.EntityKey.EntityKeyValues[0].Value;
}
public static string GetFullMessage(this Exception ex)
{
return ex.InnerException == null ? ex.Message : $"{ex.Message}\n{ex.InnerException.GetFullMessage()}";
}
public static string Replace(this string source, string oldString, string newString, StringComparison comp)
{
int index = source.IndexOf(oldString, comp);
if (index >= 0)
{
source = source.Remove(index, oldString.Length);
source = source.Insert(index, newString);
}
if (source.IndexOf(oldString, comp) != -1)
source = Replace(source, oldString, newString, comp);
return source;
}
}
I am using an IDbCommandTreeInterceptor to enable soft deletes on my model.
System.Data.Entity.Infrastructure.Interception.DbInterception.Add(
new SoftDeleteInterception());
I want to be able to disable the interceptor temporarily so that I can select a "deleted" entity for auditing purposes.
However, It seems like the DbInterception collection is assembly-wide.
Is there any way to create a new DbContext without interception on?
Or even a way to add the interceptor to the DbContext every time it is created?
I have extended my db context class with additional property
[DbConfigurationType(typeof(DbConfig))]
public partial class YourEntitiesDB
{
public bool IgnoreSoftDelete { get; set; }
}
Then in the TreeCreated(...) method i check this flag and if true then it just doesn't go further to the QueryVisitor
public class SoftDeleteInterceptor : IDbCommandTreeInterceptor
{
public SoftDeleteInterceptor()
{
}
public void TreeCreated(DbCommandTreeInterceptionContext interceptionContext)
{
var db = interceptionContext.DbContexts.FirstOrDefault() as YourEntitiesDB;
if (db!=null && db.IgnoreSoftDelete)
{
// Ignore soft delete interseptor (Used in archives)
return;
}
if (interceptionContext.OriginalResult.DataSpace == DataSpace.CSpace)
{
var queryCommand = interceptionContext.Result as DbQueryCommandTree;
if (queryCommand != null)
{
var newQuery = queryCommand.Query.Accept(new SoftDeleteQueryVisitor());
interceptionContext.Result = new DbQueryCommandTree(
queryCommand.MetadataWorkspace,
queryCommand.DataSpace,
newQuery);
}
var deleteCommand = interceptionContext.OriginalResult as DbDeleteCommandTree;
if (deleteCommand != null)
{
var column = SoftDeleteAttribute.GetSoftDeleteColumnName(deleteCommand.Target.VariableType.EdmType);
if (column != null)
{
var setClauses = new List<DbModificationClause>();
var table = (EntityType)deleteCommand.Target.VariableType.EdmType;
if (table.Properties.Any(p => p.Name == column))
{
setClauses.Add(DbExpressionBuilder.SetClause(
DbExpressionBuilder.Property(
DbExpressionBuilder.Variable(deleteCommand.Target.VariableType, deleteCommand.Target.VariableName),
column),
DbExpression.FromBoolean(true)));
}
var update = new DbUpdateCommandTree(
deleteCommand.MetadataWorkspace,
deleteCommand.DataSpace,
deleteCommand.Target,
deleteCommand.Predicate,
setClauses.AsReadOnly(),
null);
interceptionContext.Result = update;
}
}
}
}
}
In order to use it i just set the flag to true when needed
YuorEntitiesDB DB = new YuorEntitiesDB();
DB.IgnoreSoftDelete = true;
DB.Records.Where(...)
In my programme I've a large array of strings(say 1600) which I want to show as a CheckBox list. The array is actually the location of all the songs in one's PC, and thus can gradually be bigger. I don't wanna use ListView<String> as the CheckBox list is more efficient and above all visually better for my purpose. I'm currently doing the below :
private void listAll() {
songs = MediaManager.getAllSongs();
VBox vb = new VBox();
vb.setSpacing(5);
vb.getStyleClass().add("background");
if (songs != null) {
Service s = new Service() {
#Override
protected Task createTask() {
Task t = new Task() {
#Override
protected Object call() throws Exception {
for (String song : songs) {
addSong(song, vb);
c++;
updateMessage(c+" songs");
}
return null;
}
};
t.messageProperty().addListener((obs,o,n)->{
count.setText(n);
});
return t;
}
};
s.start();
ScrollPane sp = new ScrollPane(vb);
getChildren().add(sp);
}
}
private void addSong(String n, Pane p) {
String toAdd = "";
Media m = new Media(Paths.get(n).toUri().toString());
if (m.getMetadata().get("title") == null || !(m.getMetadata().get("title").equals(""))) {
toAdd = m.getSource().split("/")[m.getSource().split("/").length - 1].replace("%20", " ").replace(".mp3", "");
} else {
toAdd = ((String) m.getMetadata().get("title"));
}
SongBox s = new SongBox(toAdd);
s.setUserData(n);
p.getChildren().add(s);
}
class SongBox extends CheckBox {
public SongBox(String t) {
this();
setText(t);
}
public SongBox() {
super();
setOnAction((ActionEvent evt) -> {
if (isSelected()) {
if (!playNow.isVisible()) {
playNow.setVisible(true);
}
path = (String) getUserData();
selected.add((String) getUserData());
} else {
selected.remove((String) getUserData());
if (selected.size() == 0) {
playNow.setVisible(false);
}
}
});
}
}
First of all, that is not showing the complete array. Whenever I'm going back and returning to it, the number of songs get changed. Secondly, the whole UI is getting sluggish(sometimes also hanging my PC). Moreover, I can't cancel the Service when I've gone to the previous window, as it's always returning false. Anyone have a better approach?
I am working on a 'proof of concept' Silverlight 4 project and am learning the way of THE ASYNC. I have stopped fighting the urge to implement some pseudo-synchronous smoke and mirrors technique. I am going to learn to stop worrying and love THE ASYNC.
Most of the time I just use a BusyIndicator while async methods are running and all is good but I have run into a few situations where I need to call methods sequentially. I put together this example and it works. But in my experience... if it works... there is something wrong with it.
When is this going to blow up in my face or steal my wife or date one of my daughters?
Is there a better way to do this?
The Code:
public class CustomPage : Page
{
static readonly object _AsyncMethodChain_Lock = new object();
private Dictionary<Action<object>, string> _AsyncMethodChain = new Dictionary<Action<object>, string>();
public Dictionary<Action<object>, string> AsyncMethodChain
{
get { lock (_AsyncMethodChain_Lock) { return this._AsyncMethodChain; } }
set { lock (_AsyncMethodChain_Lock) { this._AsyncMethodChain = value; } }
}
private void CustomPage_Loaded(object sender, RoutedEventArgs e)
{
if (!System.ComponentModel.DesignerProperties.GetIsInDesignMode(this))
{
var user = this.SecurityProvider.UserObject as TimeKeeper.UserServiceReference.User;
if (user == null)
return;
this.AsyncMethodChain.Add(
data =>
{
var userServiceClient = new UserServiceClient();
userServiceClient.GetCompleted +=
(send, arg) =>
{
var userViewSource = this.Resources["userViewSource"] as CollectionViewSource;
userViewSource.Source = new List<UserServiceReference.User>(new UserServiceReference.User[1] { arg.Result });
userViewSource.View.MoveCurrentToPosition(0);
this.AsyncMethodChain.ExecuteNext(arg.Result.UserID, this.BusyIndicator);
};
userServiceClient.GetAsync(user.UserID);
},
"Loading user..."
);
this.AsyncMethodChain.Add(
data =>
{
var userID = (int)data;
var timeLogServiceClient = new TimeLogServiceClient();
timeLogServiceClient.FindByUserIDCompleted +=
(send, arg) =>
{
var timeLogViewSource = this.Resources["timeLogViewSource"] as CollectionViewSource;
timeLogViewSource.Source = arg.Result;
this.AsyncMethodChain.ExecuteNext(null, this.BusyIndicator);
};
timeLogServiceClient.FindByUserIDAsync(userID);
},
"Loading time logs..."
);
this.AsyncMethodChain.ExecuteNext(null, this.BusyIndicator);
}
}
}
public static class Extensions
{
public static void ExecuteNext(this Dictionary<Action<object>, string> methods, object data, BusyIndicator busyIndicator)
{
if (methods.Count <= 0)
{
busyIndicator.BusyContent = "";
busyIndicator.IsBusy = false;
return;
}
else
{
var method = methods.Keys.ToList<Action<object>>()[0];
busyIndicator.BusyContent = methods[method];
busyIndicator.IsBusy = true;
}
methods.ExecuteNext(data);
}
public static void ExecuteNext(this Dictionary<Action<object>, string> methods, object data)
{
var method = methods.Keys.ToList<Action<object>>()[0];
methods.Remove(method);
method(data);
}
}
What you have sine looks pretty good, but if you are still worried about the callsequence i would sugest that you create a new method in your webservice which will call the other four in whatever sequence you need them to and call this new method from your silverlight application.
I dont see the need for you to do that as your current implemenation is pretty good as well.