I'd like to log some messages only once while the program is running or be able to "unlock" the "unique" message so that it would be logged again.
Of course this can be done on code keeping track of the "unique" message and calling the log method only once, but I wonder if it is possible to do so with some configuration trick.
In pseudo code:
DO
i = i++
log.info("this should be unique")
log.info("this can be repeated")
LOOP WHILE i < 10
unlock("this should be unique")
log.info("this should be unique")
log.info("this can be repeated")
After that I should see "this should be unique" 2 times on my log.
Your other question made me consider this one also. It should be possible to create a system to log any message only once (before reset) by using any buffering appender and a custom ITriggeringEventEvaluator.
Set the buffer size to 0 and the Evaluator to the custom one. In the custom Evaluator, keep track of all messages that went through the IsTriggeringEvent method and filter out the ones that have already been processed.
It is possible to configure the messages that would be filtered, however you want to do it (config file, DB, etc), and how they can be logged again (timeout of filtering, explicit reset, etc) through the evaluator code.
Related
I have a simple Kafka streams scenario where I am doing a groupyByKey then reduce and then an action. There could be duplicate events in the source topic hence the groupyByKey and reduce
The action could error and in that case, I need the streams app to reprocess that event. In the example below I'm always throwing an error to demonstrate the point.
It is very important that the action only ever happens once and at least once.
The problem I'm finding is that when the streams app reprocesses the event, the reduce function is being called and as it returns null the action doesn't get recalled.
As only one event is produced to the source topic TOPIC_NAME I would expect the reduce to not have any values and skip down to the mapValues.
val topologyBuilder = StreamsBuilder()
topologyBuilder.stream(
TOPIC_NAME,
Consumed.with(Serdes.String(), EventSerde())
)
.groupByKey(Grouped.with(Serdes.String(), EventSerde()))
.reduce { current, _ ->
println("reduce hit")
null
}
.mapValues { _, v ->
println(Id: "${v.correlationId}")
throw Exception("simulate error")
}
To cause the issue I run the streams app twice. This is the output:
First run
Id: 90e6aefb-8763-4861-8d82-1304a6b5654e
11:10:52.320 [test-app-dcea4eb1-a58f-4a30-905f-46dad446b31e-StreamThread-1] ERROR org.apache.kafka.streams.KafkaStreams - stream-client [test-app-dcea4eb1-a58f-4a30-905f-46dad446b31e] All stream threads have died. The instance will be in error state and should be closed.
Second run
reduce hit
As you can see the .mapValues doesn't get called on the second run even though it errored on the first run causing the streams app to reprocess the same event again.
Is it possible to be able to have a streams app re-process an event with a reduced step where it's treating the event like it's never seen before? - Or is there a better approach to how I'm doing this?
I was missing a property setting for the streams app.
props["processing.guarantee"]= "exactly_once"
By setting this, it will guarantee that any state created from the point of picking up the event will rollback in case of a exception being thrown and the streams app crashing.
The problem was that the streams app would pick up the event again to re-process but the reducer step had state which has persisted. By enabling the exactly_once setting it ensures that the reducer state is also rolled back.
It now successfully re-processes the event as if it had never seen it before
I´ve got a Job that runs a TaskletStep, then a chunk-based step and then another TaskletStep.
In each of these steps, errors (in the form of Exceptions) can occur.
The chunk-based step looks like this:
stepBuilderFactory
.get("step2")
.chunk<SomeItem, SomeItem>(1)
.reader(flatFileItemReader)
.processor(itemProcessor)
.writer {}
.faultTolerant()
.skipPolicy { _ , _ -> true } // skip all Exceptions and continue
.taskExecutor(taskExecutor)
.throttleLimit(taskExecutor.corePoolSize)
.build()
The whole job definition:
jobBuilderFactory.get("job1")
.validator(validator())
.preventRestart()
.start(taskletStep1)
.next(step2)
.next(taskletStep2)
.build()
I expected that Spring Batch somehow picks up the Exceptions that occur along the way, so I can then create a Report including them after the Job has finished processing. Looking at the different contexts, there´s also fields that should contain failureExceptions. However, it seems there´s no such information (especially for the chunked step).
What would be a good approach if I need information about:
what Exceptions did occur in which Job execution
which Item was the one that triggered it
The JobExecution provides a method to get all failure exceptions that happened during the job. You can use that in a JobExecutionListener#afterJob(JobExecution jobExecution) to generate your report.
In regards to which items caused the issue, this will depend on where the exception happens (during the read, process or write operation). For this requirement, you can use one of the ItemReadListener, ItemProcessListener or ItemWriteListener to keep record of the those items (For example, by adding them to the job execution context to be able to get access to them in the JobExecutionListener#afterJob method for your report).
I am using the taskqueue API to send multiple emails is small groups with mailgun. My code looks more or less like this:
class CpMsg(ndb.Model):
group = ndb.KeyProperty()
sent = ndb.BooleanProperty()
#Other properties
def send_mail(messages):
"""Sends a request to mailgun's API"""
# Some code
pass
class MailTask(TaskHandler):
def post(self):
p_key = utils.key_from_string(self.request.get('p'))
msgs = CpMsg.query(
CpMsg.group==p_key,
CpMsg.sent==False).fetch(BATCH_SIZE)
if msgs:
send_mail(msgs)
for msg in msgs:
msg.sent = True
ndb.put_multi(msgs)
#Call the task again in COOLDOWN seconds
The code above has been working fine, but according to the docs, the taskqueue API guarantees that a task is delivered at least once, so tasks should be idempotent. Now, most of the time this would be the case with the above code, since it only gets messages that have the 'sent' property equal to False. The problem is that non ancestor ndb queries are only eventually consistent, which means that if the task is executed twice in quick succession the query may return stale results and include the messages that were just sent.
I thought of including an ancestor for the messages, but since the sent emails will be in the thousands I'm worried that may mean having large entity groups, which have a limited write throughput.
Should I use an ancestor to make the queries? Or maybe there is a way to configure mailgun to avoid sending the same email twice? Should I just accept the risk that in some rare cases a few emails may be sent more than once?
One possible approach to avoid the eventual consistency hurdle is to make the query a keys_only one, then iterate through the message keys to get the actual messages by key lookup (strong consistency), check if msg.sent is True and skip sending those messages in such case. Something along these lines:
msg_keys = CpMsg.query(
CpMsg.group==p_key,
CpMsg.sent==False).fetch(BATCH_SIZE, keys_only=True)
if not msg_keys:
return
msgs = ndb.get_multi(msg_keys)
msgs_to_send = []
for msg in msgs:
if not msg.sent:
msgs_to_send.append(msg)
if msgs_to_send:
send_mail(msgs_to_send)
for msg in msgs_to_send:
msg.sent = True
ndb.put_multi(msgs_to_send)
You'd also have to make your post call transactional (with the #ndb.transactional() decorator).
This should address the duplicates caused by the query eventual consistency. However there still is room for duplicates caused by transaction retries due to datastore contention (or any other reason) - as the send_mail() call isn't idempotent. Sending one message at a time (maybe using the task queue) could reduce the chance of that happening. See also GAE/P: Transaction safety with API calls
I'm using NServiceBus to handle some calculation messages. I have a new requirement to handle calculation errors by writing them the same database. I'm using NHibernate as my DAL which auto enlists to the NServiceBus transaction and provides rollback in case of exceptions, which is working really well. However if I write this particular error to the database, it is also rolled back which is a problem.
I knew this would be a problem, but I thought I could just wrap the call in a new transaction with the TransactionScopeOption = Suppress. However the error data is still rolled back. I believe that's because it was using the existing session with has already enlisted in the NServiceBus transaction.
Next I tried opening a new session from the existing SessionFactory within the suppression transaction scope. However the first call to the database to retrieve or save data using this new session blocks and then times out.
InnerException: System.Data.SqlClient.SqlException
Message=Timeout expired. The timeout period elapsed prior to completion of the >operation or the server is not responding.
Finally I tried creating a new SessionFactory using it to open a new session within the suppression transaction scope. However again it blocks and times out.
I feel like I'm missing something obvious here, and would greatly appreciate any suggestions on this probably common task.
As Adam suggests in the comments, in most cases it is preferred to let the entire message fail processing, giving the built-in Retry mechanism a chance to get it right, and eventually going to the error queue. Then another process can monitor the error queue and do any required notification, including logging to a database.
However, there are some use cases where the entire message is not a failure, i.e. on the whole, it "succeeds" (whatever the business-dependent definition of that is) but there is some small part that is in error. For example, a financial calculation in which the processing "succeeds" but some human element of the data is "in error". In this case I would suggest catching that exception and sending a new message which, when processed by another endpoint, will log the information to your database.
I could see another case where you want the entire message to fail, but you want the fact that it was attempted noted somehow. This may be closest to what you are describing. In this case, create a new TransactionScope with TransactionScopeOption = Suppress, and then (again) send a new message inside that scope. That message will be sent whether or not your full message transaction rolls back.
You are correct that your transaction is rolling back because the NHibernate session is opened while the transaction is in force. Trying to open a new session inside the suppressed transaction can cause a problem with locking. That's why, most of the time, sending a new message asynchronously is part of the solution in these cases, but how you do it is dependent upon your specific business requirements.
I know I'm late to the party, but as an alternative suggestion, you coudl simply raise another separate log message, which NSB handles independently, for example:
public void Handle(DebitAccountMessage message)
{
var account = this.dbcontext.GetById(message.Id);
if (account.Balance <= 0)
{
// log request - new handler
this.Bus.Send(new DebitAccountLogMessage
{
originalMessage = message,
account = account,
timeStamp = DateTime.UtcNow
});
// throw error - NSB will handle
throw new DebitException("Not enough funds");
}
}
public void Handle(DebitAccountLogMessage message)
{
var messageString = message.originalMessage.Dump();
var accountString = message.account.Dump(DumpOptions.SuppressSecurityTokens);
this.Logger.Log(message.UniqueId, string.Format("{0}, {1}", messageString, accountString);
}
How do you flush the io buffer in Erlang?
For instance:
> io:format("hello"),
> io:format(user, "hello").
This post seems to indicate that there is no clean solution.
Is there a better solution than in that post?
Sadly other than properly implementing a flush "command" in the io/kernel subsystems and making sure that the low level drivers that implement the actual io support such a command you really have to simply rely on the system quiescing before closing. A failing I think.
Have a look at io.erl/io_lib.erl in stdlib and file_io_server.erl/prim_file.erl in kernel for the gory details.
As an example, in file_io_server (which effectively takes the request from io/io_lib and routes it to the correct driver), the command types are:
{put_chars,Chars}
{get_until,...}
{get_chars,...}
{get_line,...}
{setopts, ...}
(i.e. no flush)!
As an alternative you could of course always close your output (which would force a flush) after every write. A logging module I have does something like this every time and it doesn't appear to be that slow (it's a gen_server with the logging received via cast messages):
case file:open(LogFile, [append]) of
{ok, IODevice} ->
io:fwrite(IODevice, "~n~2..0B ~2..0B ~4..0B, ~2..0B:~2..0B:~2..0B: ~-8s : ~-20s : ~12w : ",
[Day, Month, Year, Hour, Minute, Second, Priority, Module, Pid]),
io:fwrite(IODevice, Msg, Params),
io:fwrite(IODevice, "~c", [13]),
file:close(IODevice);
io:put_chars(<<>>)
at the end of the script works for me.
you could run
flush().
from the shell, or try
flush()->
receive
_ -> flush()
after 0 -> ok
end.
That works more or less like a C flush.