I would like to implement a custom endpoint class to check Zookeeper health:
http://localhost:8080/actuator/health/zookeeper
PROBLEM: Do I extend AbstractHealthIndicaitor or implement HealthIndicator class?
HealthIndicator class
public class CustomHealth implements HealthIndicator {
#Override
public Health health() {
int errorCode = check(); // perform some specific health check
if (errorCode != 0) {
return Health.down()
.withDetail("Error Code", errorCode).build();
}
return Health.up().build();
}
public int check() {
// Our logic to check zookeeper health
return 0;
}
}
AbstractHealthIndicator class
public class CustomHealth extends AbstractHealthIndicator {
#Override
protected void doHealthCheck(Health.Builder builder) throws Exception
{
// Our logic to check zookeeper health
}
}
I'm confused on which approach to use. I believe the logic to check zookeeper health is to simply declare an CuratorFramework object then do curator.getState() and return builder from there and for endpoint, attach #RestControllerEndPoint to declare the path. Please help!
It is up to you which one to choose, the difference is that AbstractHealthIndicator:
Provides you with the Health.Builder instance so you don't need to create one manually
Wraps doHealthCheck(builder) call with try-catch, that returns status DOWN if your healthcheck has failed with exception.
So in general the AbstractHealthIndicator is more convenient to use as you can skip error handling. Choose implementing raw HealthIndicator when you need to provide custom status details on exception.
For example of Zookeeper Health Indicator please refer to existing one provided with spring-cloud-zookeeper https://github.com/spring-cloud/spring-cloud-zookeeper/blob/master/spring-cloud-zookeeper-core/src/main/java/org/springframework/cloud/zookeeper/ZookeeperHealthIndicator.java
Regarding the endpoint /actuator/health/zookeeper, I suggest you to use new feature introduced in SpringBoot 2.2.0 called Health Indicator Groups https://spring.io/blog/2019/10/16/spring-boot-2-2-0#health-indicator-groups
In short, if you use component scan and named custom health indicator MyZookeeperHealthIndicator, then add following properties to register it under custom zookeeper group:
management.endpoint.health.group.zookeeper.include=myZookeeper
After that, custom health details will be displayed under myZookeeper component at /actuator/health/zookeeper
Check following docs for more information:
Writing Custom HealthIndicators https://docs.spring.io/spring-boot/docs/current/reference/html/production-ready-features.html#writing-custom-healthindicators
Health Groups https://docs.spring.io/spring-boot/docs/current/reference/html/production-ready-features.html#health-groups
I have been developing a system which is heavily using stateless service fabric services. I thought I had a good idea how things worked however I am doing something slightly different now and it is obvious my understanding is limited.
I have a cluster of 5 nodes and all my services have an instance count of -1 for simplicity currently. With everything on every node it means I can watch a single node for basic behavior correctness.
I just added a new service which needs an instance count of 1. However it seems I am unable to resolve this service correctly. Instead SF tries to resolve the service on each machine which fails for all machines except the one where the single service exists.
It was my assumption that SF would automagically resolve a reference to a service anywhere in the cluster. If that reference fails then it would automagically resolve a new reference and so on. It appears that this is not correct at least for the way I am currently doing things.
I can find an instance using code similar to this however what happens if that instance fails. How do I get another reference?
I can resolve for every call like this however that seems like a terrible idea when I really only want to resolve a IXyzService and pass that along.
This is how I am resolving services since I am using the V2 custom serialization.
var _proxyFactory = new ServiceProxyFactory(c =>
{
return new FabricTransportServiceRemotingClientFactory(
serializationProvider: new CustomRemotingSerializationProvider(Logger)
);
});
var location = new Uri("fabric:/xyz/abcService");
var proxy = _proxyFactory.CreateServiceProxy<TService>(location);
This does actually work however it appears to only resolve a service on the same machine. So ServiceA would resolve a reference to ServiceB on the same machine. However if ServiceB doesn't exist on the machine for a valid reason then the resolution would fail.
Summary:
What is the correct way for ServiceA to use the V2 custom serialization ServiceProxyFactory to resolve an interface reference to ServiceB wherever ServiceA and ServiceB are in the cluster?
Update:
The evidence it doesn't work is the call to resolve hangs forever. According to this link that is correct behavior because the service will eventually come up. However only 1 node ever resolved it correctly and that is the node where the single instance is active. I have tried several things even waiting 30 seconds just to make sure it wasn't an init issue.
var proxy = _proxyFactory.CreateServiceProxy<TService>(location);
// Never gets here except on the one node.
SomethingElse(proxy);
Listener code
This essentially follows the V2 custom serialization tutorial almost exactly.
var listeners = new[]
{
new ServiceInstanceListener((c) =>
{
return new FabricTransportServiceRemotingListener(c, this, null,
new CustomRemotingSerializationProvider(Logger));
})
};
public class HyperspaceRemotingSerializationProvider : IServiceRemotingMessageSerializationProvider
{
#region Private Variables
private readonly ILogger _logger;
private readonly Action<RequestInfo> _requestAction;
private readonly Action<RequestInfo> _responseAction;
#endregion Private Variables
public CustomRemotingSerializationProvider(ILogger logger, Action<RequestInfo> requestAction = null, Action<RequestInfo> responseAction = null)
{
_logger = logger;
_requestAction = requestAction;
_responseAction = responseAction;
}
public IServiceRemotingRequestMessageBodySerializer CreateRequestMessageSerializer(Type serviceInterfaceType, IEnumerable<Type> requestWrappedTypes,
IEnumerable<Type> requestBodyTypes = null)
{
return new RequestMessageBodySerializer(_requestAction);
}
public IServiceRemotingResponseMessageBodySerializer CreateResponseMessageSerializer(Type serviceInterfaceType, IEnumerable<Type> responseWrappedTypes,
IEnumerable<Type> responseBodyTypes = null)
{
return new ResponseMessageBodySerializer(_responseAction);
}
public IServiceRemotingMessageBodyFactory CreateMessageBodyFactory()
{
return new MessageBodyFactory();
}
}
Connection code
_proxyFactory = new ServiceProxyFactory(c =>
{
return new FabricTransportServiceRemotingClientFactory(
serializationProvider: new CustomRemotingSerializationProvider(Logger)
);
});
// Hangs here - tried different partition keys or not specifying one.
var proxy = _proxyFactory.CreateServiceProxy<TService>(location, ServicePartitionKey.Singleton);
I'm using Spring cloud starter gateway 2.0.1.RELEASE along with Starter netflix hystrix. Is it possible to provide a HystrixCommand in route definition like below?.
builder.routes()
.route( r -> r.path("path")
.and().method(HttpMethod.GET)
.filters(f -> f.hystrix("myHystrixCommandName"))
.uri("/destination")
.id("route_1"));
My goal is to execute a fallback method without forwarding the request to a fallback uri.
Also I cannot use static fallback uri option as I need path params and request params to determine the fallback response. Any help is highly appreciated!.
I faced the same issue. This is how I solved it:
First, These are the routes:
builder.routes()
.route( r -> r.path("/api/{appid}/users/{userid}")
.and().method(HttpMethod.GET)
.filters(f -> f.hystrix("myHystrixCommandName")
.setFallbackUri("forward:/hystrixfallback"))
.uri("/destination")
.id("route_1"));
The path predicate processes the url and extracts the path variables, those are saved into ServerWebExchange.getAttributes(), with a key defined as PathRoutePredicate.URL_PREDICATE_VARS_ATTR or ServerWebExchangeUtils.URI_TEMPLATE_VARIABLES_ATTRIBUTE
You can find more information about path predicate here
Second, I created a #RestController to handle the forward from Hystrix injecting the ServerWebExchange:
#RestController
#RequestMapping("/hystrixfallback")
public class ServiceFallBack {
#RequestMapping(method = {RequestMethod.GET})
public String get(ServerWebExchange serverWebExchange) {
//Get the PathMatchInfo
PathPattern.PathMatchInfo pathMatchInfo = serverWebExchange.getAttribute(ServerWebExchangeUtils.URI_TEMPLATE_VARIABLES_ATTRIBUTE);
//Get the template variables
Map<String, String> urlTemplateVariables = pathMatchInfo.getUriVariables();
//TODO Logic using path variables
return "result";
}
}
In the pre-GA version of Service Fabric I was registering a custom serializer like this:
protected override IReliableStateManager CreateReliableStateManager()
{
IReliableStateManager result = new ReliableStateManager(
new ReliableStateManagerConfiguration(
onInitializeStateSerializersEvent: InitializeStateSerializers));
return result;
}
private Task InitializeStateSerializers()
{
StateManager.TryAddStateSerializer(new KFOBinarySerializer());
return Task.FromResult(false);
}
However, the CreateReliableStateManager method was removed in the GA version. I've struggled to get something working in its place. Currently I'm calling
StateManager.TryAddStateSerializer(new KFOBinarySerializer());
from within the service's RunAsync method, which appears to work fine.
What is the recommended way to register a custom serializer?
TryAddStateSerializer is deprecated. Anyone know if this is because custom serialization support will go away or if it will simply be supported through some other mechanism?
You can create the state manager in the StatefulService's constructor (full example here):
class MyService : StatefulService
{
public MyService(StatefulServiceContext serviceContext)
: base(serviceContext, CreateReliableStateManager()) { }
private static IReliableStateManager CreateReliableStateManager() { ... }
}
Regarding the deprecated API, Microsoft says it's safe to use, but it will change in the future.
How to restrict a CQ5/Custom component to add only once per page.? I want to restrict the drag and drop of component into the page when the author is going to add the same component for the second time into the same page.
One option is to include the component directly in the JSP of the template and exclude it from the list of available components in the sidekick. To do so, add the component directly to your JSP (foundation carousel in this example):
<cq:include path="carousel" resourceType="foundation/components/carousel" />
To hide the component from the sidekick, either set:
componentGroup: .hidden
or exclude it from the list of "Allowed Components" using design mode.
If you need to allow users to create a page without this component you can provide a second template with the cq:include omitted.
Thanks Rampant, I have followed your method and link stated.
Posting link again : please follow this blog
It was really helpful. I am posting the implementation whatever I have done.
It worked fine for me. One can definitely improve the code quality, this is raw code and is just for reference.
1.Servlet Filter
Keep this in mind that,if any resource gets refereshed, this filter will execute. So you need to filter the contents at your end for further processing.
P.S. chain.doFilter(request,response); is must. or cq will get hanged and nothing will be displayed.
#SlingFilter(generateComponent = false, generateService = true, order = -700,
scope = SlingFilterScope.REQUEST)
#Component(immediate = true, metatype = false)
public class ComponentRestrictorFilter implements Filter {
public void init(FilterConfig filterConfig) throws ServletException {}
#Reference
private ResourceResolverFactory resolverFactory;
public void doFilter(ServletRequest request, ServletResponse response, FilterChain chain)
throws IOException, ServletException {
WCMMode mode = WCMMode.fromRequest(request);
if (mode == WCMMode.EDIT) {
SlingHttpServletRequest slingRequest = (SlingHttpServletRequest) request;
PageManager pageManager = slingRequest.getResource().getResourceResolver().adaptTo(PageManager.class);
Page currentPage = pageManager.getContainingPage(slingRequest.getResource());
logger.error("***mode" + mode);
if (currentPage != null )) {
ComponentRestrictor restrictor = new ComponentRestrictor(currentPage.getPath(), RESTRICTED_COMPONENT);
restrictor.removeDuplicateEntry(resolverFactory,pageManager);
}
chain.doFilter(request, response);
}
}
public void destroy() {}
}
2.ComponentRestrictor class
public class ComponentRestrictor {
private String targetPage;
private String component;
private Pattern pattern;
private Set<Resource> duplicateResource = new HashSet<Resource>();
private Logger logger = LoggerFactory.getLogger(ComponentRestrictor.class);
private Resource resource = null;
private ResourceResolver resourceResolver = null;
private ComponentRestrictorHelper helper = new ComponentRestrictorHelper();
public ComponentRestrictor(String targetPage_, String component_){
targetPage = targetPage_ + "/jcr:content";
component = component_;
}
public void removeDuplicateEntry(ResourceResolverFactory resolverFactory, PageManager pageManager) {
pattern = Pattern.compile("([\"']|^)(" + component + ")(\\S|$)");
findReference(resolverFactory, pageManager);
}
private void findReference(ResourceResolverFactory resolverFactory, PageManager pageManager) {
try {
resourceResolver = resolverFactory.getAdministrativeResourceResolver(null);
resource = resourceResolver.getResource(this.targetPage);
if (resource == null)
return;
search(resource);
helper.removeDuplicateResource(pageManager,duplicateResource);
} catch (LoginException e) {
logger.error("Exception while getting the ResourceResolver " + e.getMessage());
}
resourceResolver.close();
}
private void search(Resource parentResource) {
searchReferencesInContent(parentResource);
for (Iterator<Resource> iter = parentResource.listChildren(); iter.hasNext();) {
Resource child = iter.next();
search(child);
}
}
private void searchReferencesInContent(Resource resource) {
ValueMap map = ResourceUtil.getValueMap(resource);
for (String key : map.keySet()) {
if (!helper.checkKey(key)) {
continue;
}
String[] values = map.get(key, new String[0]);
for (String value : values) {
if (pattern.matcher(value).find()) {
logger.error("resource**" + resource.getPath());
duplicateResource.add(resource);
}
}
}
}
}
3.To remove the node/ resource
Whichever resource you want to remove/delete just use PageManager api
pageManeger.delete(resource,false);
That's it !!! You are good to go.
None of the options looks easy to implement. The best approach I found is to use the ACS Commons Implementation which is very easy and can be adopted into any project.
Here is the link and how to configure it:
https://github.com/Adobe-Consulting-Services/acs-aem-commons/pull/639
Enjoy coding !!!
you can't prevent that without doing some massive hacking to the ui code, and even then, you've only prevented it from one aspect of the ui. there's still crxde, and then the ability to POST content.
if this is truly a requirement, the best approach might be the following:
have the component check for a special value in the pageContext object (use REQUEST_SCOPE)
if value is not found, render component and set value
otherwise, print out a message that component can only be used once
note that you can't prevent a dialog from showing, but at the very least the author has an indication that that particular component can only be used once.
It sounds like there needs to be clarification of requirements (and understanding why).
If the authors can be trained, let them manage limits of components through authoring and review workflows.
If there is just 1 fixed location the component can appear, then the page component should include the content component, and let the component have an "enable" toggle property to determine if it should render anything. The component's group should be .hidden to prevent dragging from the sidekick.
If there is a fixed set of locations for the component, the page component can have a dropdown of the list of locations (including "none"). The page render component would then conditionally include the component in the correct location. Again, prevent dragging the component from the sidekick.
In the "hard to imagine" case that the component can appear anywhere on the page, added by authors, but limited to only 1 instance - use a wrapper component to manage including the (undraggable) component. Let the authors drag the wrapper on the page as many times as they want, but the wrapper should query the page's resources and determine if it is the first instance, and if so, include the end component. Otherwise, the wrapper does nothing.
In our experience (>2years on CQ), implementing this type of business rules via code creates a brittle solution. Also, requirements have a habit of changing. If enforced via code, development work is required instead of letting authors make changes faster & elegantly.
None of these options are that great. If you truly want a robust solution to this problem (limit the number of items on the page without hardcoding location) then the best way is with a servlet filter chain OSGI service where you can administer the number of instances and then use a resource resolver to remove offending instances.
The basic gist is:
Refresh the page on edit using cq:editConfig
Create an OSGI service implementing javax.servlet.Filter that encapsulates your business rules.
Use the filter to remove excess components according to business rules
Continue page processing.
For more details see here:
Using a servlet filter to limit the number of instances of a component per page or parsys
This approach will let you administer the number of items per page or per parsys and apply other possibly complex business rules in a way that the other offered solutions simply cannot.