Re: [support] ignite tuning help

2024-06-21 Thread Jeremy McMillan
Also, I didn't look at your network trace screen cap, but you should have zero TCP retransmissions if you set your initial TCP window send window small enough. https://www.auvik.com/franklyit/blog/tcp-window-size/ On Fri, Jun 21, 2024, 07:53 Jeremy McMillan wrote: > It could be netw

Re: [support] ignite tuning help

2024-06-21 Thread Jeremy McMillan
It could be network or persistent storage. What's the proportion of fast to slow gets? On Thu, Jun 20, 2024, 22:48 f cad wrote: > here is a screenshot example > [image: image.png] > > f cad 于2024年6月21日周五 11:45写道: > >> Hello, community: >> >> I have a cluster ,with three nodes. >> I have two

Re: Best way to update and organize nodes

2024-05-30 Thread Jeremy McMillan
This could work if you set up availability zones and use backup filters. Then you could perform maintenance one entire AZ at a time. When running during maintenance, your workload might exceed the capacity of the fraction of server nodes remaining up, so beware that. On Thu, May 30, 2024, 11:30

Re: Node requires maintenance, non-empty set of maintainance tasks is found - node is not coming up

2024-05-29 Thread Jeremy McMillan
If backup partitions are available when a node is lost, we should not expect lost partitions. There is a lot more to this story than this thread explains, so for the community: please don't follow this procedure. https://ignite.apache.org/docs/latest/configuring-caches/partition-loss-policy "A

Re: Possible too long JVM pause - Ignite 2.10

2024-05-09 Thread Jeremy McMillan
Finding happiness is unfortunately never quite that simple. 1. Understand why the garbage collector cannot function with shorter pauses. (may require GC logging configuration to provide key details) 2. Identify priorities. (ie. absolute minimal GC pauses for best latency

Re: Turning off deadlock detection

2024-03-24 Thread Jeremy McMillan
https://ignite.apache.org/docs/latest/key-value-api/transactions#deadlock-detection The property you're asking about is for diagnostics to enable prevention of future deadlocks. Turning it off is fine if you already know or can find out another way what is deadlocked and why. On Sat, Mar 23,

Re: Data loss in an Ignite application

2024-02-21 Thread Jeremy McMillan
First, logging should be configured to at least WARN level if not INFO. Ignite manages data internally at the page level. If you see errors about pages, it is low, low level ignite problems. The next level up is partitions. Errors involving partitions are mid low level ignite problems. The next

Re: ignite + external database

2024-02-06 Thread Jeremy McMillan
2. If you want the Ignite cluster to be authoritative about caches, then you should define them in the XML configuration or deploy your servers with code which can look up the intended cache configurations and implement them. If you have specific ideas how you would like to implement this, maybe

Re: Info about time series support

2024-01-05 Thread Jeremy McMillan
To answer the OP question, maybe linear regression is sufficient for making predictions in your data. Ignite isn't really designed for exploratory data analysis, so it really helps to understand the character of your data. Linear models are usually a good place to start. Does a regression line

Re: India Scala & Big Data Job Referral

2023-12-21 Thread Jeremy McMillan
It might help if you address the Ignite community with a question about Apache Ignite or GridGain skills and experience and needs. It might demonstrate your skills if you participate in the community by answering others' questions about Apache Ignite or GridGain, as it will establish your

Re: Another replicated cache oddity

2023-11-22 Thread Jeremy McMillan
but not all, copies of the element where grid > restarts do not correct the issue. This does not feel the same though. > > Raymond. > > > > > > On Thu, Nov 23, 2023 at 6:50 AM Jeremy McMillan < > jeremy.mcmil...@gridgain.com> wrote: > >> I suspect a race condition

Re: Another replicated cache oddity

2023-11-22 Thread Jeremy McMillan
I suspect a race condition with async mode caches. This is a naive guess though, as we don't have enough details. I'll assume this is a plea for help in troubleshooting methodology and the question is really "what should we look at next?" The real answer comes from tracing the insert of element E

Re: Failed to process selector key

2023-11-14 Thread Jeremy McMillan
nio.GridDirectParser@7141a1d9, directMode=true], > GridConnectionBytesVerifyFilter], accepted=true, markedForClose=false]]] > > On the client side I don't see any errors happening around that time, I > have also searched for warnings, but nothing. > > I've seen this post where they

Re: Failed to process selector key

2023-11-13 Thread Jeremy McMillan
These errors look like something which does not speak Ignite binary protocol is connecting and sending useless stuff to your Ignite cluster. IgniteException: Invalid message type: 2057 Check the configuration of the client if the host generating this traffic is known, and check firewalls or

Re: i meet java.lang.OutOfMemoryError: Direct buffer memory

2023-08-19 Thread Jeremy McMillan
e some suggestions about how can i confirm that it is a > network problem? > Best Regards. > > Jeremy McMillan 于2023年8月18日周五 20:39写道: > >> This is most likely to happen when Ignite is fast and the network is >> slow. It's unclear what's happening when you experience this

Re: i meet java.lang.OutOfMemoryError: Direct buffer memory

2023-08-18 Thread Jeremy McMillan
This is most likely to happen when Ignite is fast and the network is slow. It's unclear what's happening when you experience this error, so how to fix it is also ambiguous. You could try increasing direct buffer memory in Java options. If that's not sufficient share more about your infrastructure

Re: Random failure while data insertion in ignite cache

2023-07-14 Thread Jeremy McMillan
It would be appropriate for you to share any code and configuration to help others recreate your system and the behavior you describe. Without, many people could solve multiple problems, demonstrably, and yet your problem is different enough that they are no help. On Fri, Jul 14, 2023, 01:58

Re: Ignite Server Node cluster

2023-07-11 Thread Jeremy McMillan
This will depend on how you are deploying your Spark workers. Whatever you are doing to control Spark workers should be replicated to control startup and shutdown of your Ignite nodes. Please start with the included ignite.sh or ignite.bat scripts found in the bin folder of your Ignite

Re: Ignite for Parquet files

2023-06-30 Thread Jeremy McMillan
Python doesn't at this time go anywhere near Ignite CacheStore. You would need to implement the CacheStore in Java or some other language which compiles to JVM runtime/jar. There's a talk from the most recent summit on using Groovy, if you want a higher level language than Java, but theoretically

Re: How communication happens when using Multicast + Static IP finder

2023-06-18 Thread Jeremy McMillan
You will need to research the documentation on three subjects: * Discovery * Data distribution * Client Cluster awareness Much of what you seem to be asking is somewhat configurable, so the answer is "it depends." Also, the question seems very broad, and you might benefit from learning

Re: Ignite thin client continuous query listener cannot listen to all events

2023-05-24 Thread Jeremy McMillan
Thanks for bringing this up! https://ignite.apache.org/docs/latest/key-value-api/continuous-queries#events-delivery-guarantees This sounds like you may have found a bug, but the details you've provided are not sufficient to help others recreate and observe it for themselves, and this effort

Re: Large data transfers with Ignite or Kafka?

2023-03-24 Thread Jeremy McMillan
That's a big question, and it isn't clear whether there's a large or small ratio of reads to writes between these microservices, for example. It isn't clear what your latency tolerance is for these large transfers either. This sounds like a big endeavor, and if there's money to be made, your best

Re: Ignite Cluster issues with larger latency between nodes

2023-03-09 Thread Jeremy McMillan
Has this kind of benchmark ever been published for any p2p cluster technology? What questions would it answer if there were such benchmarks for Ignite? Maybe this will help: There is an established algorithm for estimating the amount of buffer space necessary to keep a pipeline from stuttering

Re: Unable create a Cache SQL table on Ignite Node.

2023-03-01 Thread Jeremy McMillan
Java exception troubleshooting usually begins with an error message and a stack trace. Can we get that added to your fine description of how you found the error? We still don't know what error you found. Also please provide your config, with secrets redacted, of course. Both the details of the

Re: How to delete data of a specified partition with high performance

2023-03-01 Thread Jeremy McMillan
These documentation pages should help. https://ignite.apache.org/docs/latest/key-value-api/basic-cache-operations https://ignite.apache.org/docs/latest/configuring-caches/atomicity-modes On Tue, Feb 28, 2023 at 11:42 PM 38797715 <38797...@qq.com> wrote: > hi, > > How to delete data of a

Re: Performance of data stream on 3 cluster node.

2023-02-28 Thread Jeremy McMillan
Have you tried tracing the workload on the 100% and 40% nodes for comparison? There just isn't enough detail in your question to help predict what should be happening with the cluster workload. For a starting point, please identify your design goals. It's easy to get confused by advice that seeks

Re: random scenario of insertion operations failed.

2023-02-16 Thread Jeremy McMillan
The first step to begin debugging would be to configure logging, and reduce the Ignite pods to one. Increase the logging details until you see consistent positive indication of Ignite doing INSERT for each operation. If the failures do not appear, gradually increase the workload and then the pods

Re: How to set -DIGNITE_QUIET=false in service.sh?

2023-02-01 Thread Jeremy McMillan
This seems to be at the level of a high quality bug report, and has enough detail that a fix could probably be implemented and submitted as a PR fairly easily. Are you familiar with the contributor process? On Wed, Feb 1, 2023, 04:56 Айсина Роза Мунеровна wrote: > Hola! > We run Ignite via

Re: Safe settings for ignite cache with external store

2023-01-27 Thread Jeremy McMillan
If your problem is simple and popular, then solutions will chase you. Consider maybe you might be trying to do something really challenging that doesn't have any off the shelf solutions. You might need to approach this as a computer scientist and search for the best fit, not just best available

Re: How to check that affinity key works?

2023-01-12 Thread Jeremy McMillan
ffinity key is to enable > distributed JOIN? > > On 10 Jan 2023, at 10:48 PM, Jeremy McMillan > wrote: > > Внимание: Внешний отправитель! > Если вы не знаете отправителя - не открывайте вложения, не переходите по > ссылкам, не пересылайте письмо! > > If you are only doing col

Re: How to check that affinity key works?

2023-01-10 Thread Jeremy McMillan
If you are only doing colocated joins, then there will be no runtime overhead incurred by collecting distributed rows (colocated joins mean do not try to join data that might be distributed across nodes), so there might not be much difference in runtimes. The difference between different affinity

Re: How can I specify a column of java object in my sql select list after switching to calcite

2022-11-10 Thread Jeremy McMillan
You might want to watch the recording of the summit talk on Ignite 3 changes. There is a major change around how binary column types and objects are stored using Calcite. On Thu, Nov 10, 2022, 15:10 tore yang via user wrote: > My cache has a column of type java map, the column name is "mymap".

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-06 Thread Jeremy McMillan
ole excercise of configuring AZ as backup filter > is because we want to handle AZ level failure. > > Anyway, thanks for inputs. Will figure out further steps > > On Sun, 6 Nov 2022, 20:55 Jeremy McMillan, > wrote: > >> Don't configure 2 backups when you only have two failure

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-06 Thread Jeremy McMillan
't this mean node 3 need 2X space as compared to node 1 and node2. > Assuming backup partitions of node 3 would be equally distributed among > other two nodes. They would need almost same space. > > > On Tue, 1 Nov 2022, 23:30 Jeremy McMillan, > wrote: > >> >> >>

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-01 Thread Jeremy McMillan
er throw error if enough hardware is not present than risking >> data unavailability issue during business activity >> >> 2. Why we want 3 copies of data. It's a design choice. We want to ensure >> even if 2 nodes go down, we still have 3rd present to serve the data. >>

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-01 Thread Jeremy McMillan
e is not present than risking > data unavailability issue during business activity > > 2. Why we want 3 copies of data. It's a design choice. We want to ensure > even if 2 nodes go down, we still have 3rd present to serve the data. > > Hope I answered your question > > On Tue

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-01 Thread Jeremy McMillan
This question is a design question. What kids of fault states do you expect to tolerate? What is your failure budget? Why are you trying to make more than 2 copies of the data distribute across only two failure domains? Also "fail fast" means discover your implementation defects faster than

Re: Backup filter in ignite [Multi AZ deployment]

2022-11-01 Thread Jeremy McMillan
Using the AWS tutorial will get you a backup filter using this implementation: ClusterNodeAttributeAffinityBackupFilter There is logic to prevent a cascade of backup data onto survivor nodes in case of multiple concurrent failures if you read the documentation.

Re: Creating local cache without cluster-wide lock

2022-09-30 Thread Jeremy McMillan
Builder but noticed it creates multiple > temporary caches with the same characteristics as upstream full training > cache, i.e. distributed on all nodes. That's unnecessary in my case because > I need the subset only temporarily on the worker node. > > > > On 30.09.22 01:08, Jeremy McMill

Re: Creating local cache without cluster-wide lock

2022-09-29 Thread Jeremy McMillan
I share Stephen's curiosity about the use case. The best compromises are sensitive to situation and outcomes. Are you trying to cull training data into training, tuning, and validation subsets? Maybe there's a colocation approach that would suffice. On Thu, Sep 29, 2022, 12:26 Thomas Kramer

Re: Apache Hudi + Apache Ignite

2022-09-15 Thread Jeremy McMillan
I just read this, about hudi, and I can't see a use case for putting hudi behind an Ignite write-through cache. https://www.xenonstack.com/insights/what-is-hudi Hudi seems to be a write accelerator for Spark on HDFS, primarily. What would the expected outcome be if we assume the magic

Re: Edge Computing Read Time Application

2022-09-08 Thread Jeremy McMillan
Maybe it would be easier to start from OpenCV or something like that, figure out how to get the processing pipeline MVP for the simplest use case working, and then use Ignite as a data integration hub to scale out the architecture? My guess is the Redis example would have followed a similar

Re: distributed-computing error System.Runtime.Serialization.ISerializable

2022-08-29 Thread Jeremy McMillan
Have you followed all of the conventions in the .Net remote assembly loading doc? Have you been able to follow the example given? https://ignite.apache.org/docs/latest/net-specific/net-remote-assembly-loading On Mon, Aug 29, 2022 at 9:08 AM Charlin S wrote: > Hi, > I have started .Net node on

Re: How to enable ignite compress capability

2022-08-03 Thread Jeremy McMillan
https://github.com/apache/ignite/blob/da8a6bb4756c998aa99494d395752be96d841ec8/modules/core/src/main/java/org/apache/ignite/internal/processors/compress/FileSystemUtils.java#L45 Is Windows file storage supported for use with compression? It isn't clear whether the Java FileStore spi supports the