With these symptoms when the programmer understands the rules, there is
still a somewhat frequent bug where "cache.withKeepBinary() "is what is
required, rather than simply "cache".
On Wed, May 29, 2019 at 12:29 PM Dmitriy Pavlov wrote:
> Hi Scott,
>
> actually, users are encouraged to suggest
We had found we needed to change this in ignite.sh along these lines, so we
only had to expose one port out of the container.Otherwise you need to
expose the RMI port also.
# Newer Java versions (1.8.0_121+) allow the RMI port to be the same port.
if [ -n "$JMX_PORT" ]; then
https://apacheignite.readme.io/docs/docker-deployment
shows
sudo docker run -it --net=host -e "CONFIG_URI=$CONFIG_URI"
[-e "OPTION_LIBS=$OPTION_LIBS"]
[-e "JVM_OPTS=$JVM_OPTS"]
...
$CONFIG_URI can be https:xml which is the configuration file.
A configuration file can say to use ENVIRONMENT
We have a common spring file accessible via HTTP. Inside
and we vary the environment variables.
NOTE: the work directory has state that also needs to be persistent.
persistent copies of the above are worthless without a persistent copy of
Created https://jira.apache.org/jira/browse/IGNITE-10791 for this
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
"protocol": "tcp",
"containerPort": 9000,
"hostPort": 9000
},
{
"protocol": "tcp",
&
See attached, which we use in our AWS ECS containers.
Note that beside WAL and data, the work directory needs persistence,
because it has all the typeID mappings.
On Tue, Dec 18, 2018 at 7:32 AM Павлухин Иван wrote:
> Hi Rahul,
>
> Could you please share an ignite configuration and how do you
, Dec 3, 2018 at 3:31 PM Stanislav Lukyanov
wrote:
> Hi,
>
>
>
> Have you been able to solve this?
>
> I think specifying TcpDiscoverySpi.localAddress should work.
>
>
>
> Stan
>
>
>
> *From: *Dave Harvey
> *Sent: *17 октября 2018 г. 20:10
>
Gridgain has some kind of snapshoting add in.
You can save and restore the workDirectory from each node when the cluster
is in a stable state, provided that you use the same CONSISTENT_ID when
restoring. We were able to convert the directory name back into a
consistent ID on restore, but we had
Reverting the AWS AMI did not revert the ECS agent version.
On Fri, Oct 26, 2018 at 6:11 PM Dave Harvey wrote:
> I was running a 8x i3.8xlarge cluster on AWS ECS, and it would normally
> display
>
> ^-- H/N/C [hosts=8, nodes=8, CPUs=256]
>
>
> Then I recreated it and g
I was running a 8x i3.8xlarge cluster on AWS ECS, and it would normally display
^-- H/N/C [hosts=8, nodes=8, CPUs=256]
Then I recreated it and got this, and problems like I can no longer start visor.
^-- H/N/C [hosts=8, nodes=8, CPUs=8]
Our ECS task specifies CPU=0 -> 1 share. I'm
When we use S3 discovery and Ignite containers running under ECS using host
networking, the S3 bucket end up with 172.17.0.1#47500 along with the other
server addresses. Then on cluster startup we must wait for the network
timeout.Is there a way to avoid having this address pushed to the S3
"Ignite will only use one index per table"
I assume you mean "Ignite will only use one index per table per query"?
On Thu, Oct 11, 2018 at 1:55 PM Stanislav Lukyanov
wrote:
> Hi,
>
>
>
> It is a rather lengthy thread and I can’t dive into details right now,
>
> but AFAICS the issue now is
v
>
>
> пн, 17 сент. 2018 г. в 19:32, Dave Harvey :
>
>> I probably did not explain this clearly. When sending a message from
>> server to client using the message grid, from a context unrelated to any
>> client call, the server, as you would expect uses its install
; Hello!
>
> I think that Ignite cannot unload old version of code, unless it is loaded
> with something like URI deployment module.
> Version checking is there but server can't get rid of old code if it's on
> classpath.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
>
We have a client that uses the compute grid and message grid, as well as
the discovery API. It communicates with a server plugin. The cluster is
configured for CONTINUOUS peer class loading. In order to force the proper
code to be loaded for the compute tasks, we change the user version, e.g.,
It is my understanding that for Ignite transactions to be ACID, we need to
have the caches configured as FULL_SYNC. [ Some of the code seems to
imply only at least one of the caches in the transaction needs to be
FULL_SYNC, but that is outside the scope of my question. ]
The initial load of
It is unclear what the intended semantics of using
IgniteCompute.affinityRun() when a subset of the grid was selected. From
reading the code, my current guess is IgniteCompute.affinityRun() will run
on the primary, regardless of whether only a subset of the grid was
specified.
In my case I
at.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
> вт, 28 авг. 2018 г. в 17:37, Dave Harvey :
>
>> I did a suggested edit adding the Spring configuration of templates.
>> The rest of the current semantics seem a bit odd, so I was somewhat at a
>> loss as to wha
The DataStreamer is unordered. If you have duplicate keys with different
values, and you don't flush or take other action, then you will get an
arbitrary result. AllowOverwrite is not a solution.
Adding to the streamer returns a Future, and all of those futures are
notified when the buffer
and just always have this configuration around.
> 6) See above about the '*'.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
> сб, 25 авг. 2018 г. в 0:55, Dave Harvey :
>
>> I found what I've read in this area confusing, and here is my current
>> understandin
I found what I've read in this area confusing, and here is my current
understanding.
When creating an IgniteConfiguration in Java or XML, I can specify the
property cacheConfiguration, which is an array of CacheConfigurations.
This causes Ignite to preserve these configurations, but this will not
We are trying to load and transform a large amount of data using the
IgniteDataStreamer using a custom StreamReceiver.We'd like this to run
a lot faster, and we cannot find anything that is close to saturated,
except the data-streamer threads, queues. This is 2.5, with Ignite
persistence,
I've been able to look at cache and thread pool statistics using JVisualVM
with Mbeans support. Has anyone found a way to get these statistics out
to a tool like NewRelic or DataDog?
Thanks,
Dave Harvey
Disclaimer
The information contained in this communication from the sender
n you please elaborate with ifconfig information and stuff?
>
> Regards,
>
> --
> Ilya Kasnacheev
>
> 2018-08-03 16:53 GMT+03:00 Dave Harvey :
>
>> I've been successfully running 2.5 on AWS ECS with host or AWSVPC
>> networking for the Ignite containers. Is t
I've been successfully running 2.5 on AWS ECS with host or AWSVPC
networking for the Ignite containers. Is there any way around the fact
that with bridge networking, the Ignite node registers it's unmapped
address on S3?
Disclaimer
The information contained in this communication from the
We did the following while loading a lot of data into 2.5
1) Started data loading on 8 node cluster
2) ALTER TABLE name NOLOGGING on tables A,B,C,D but not X
3) continued loading
4) deactivated cluster
5) changed the config xml, to increase maxSize of the data region (from 2G
to 160G) and
I know Ignite only allows very limited changes to caches at runtime, e.g.,
turn on statistics or add/remove index or field.
I'm wondering if there is a way to change any of the cache configuration for
persistent caches at cluster startup.I have the impression that at some
point I saw some code
The cluster needs to agree on how to decode various versions of the
BinaryObjectSchema. Changing the type of a field name or an enum's value
are non-upwards compatible changes which Ignite cannot handle.
There is the question of the lifetime of the version of a type, and while
you may know that
I added this ticket, because we hit a similar problem, as was able to find
some quite suspect code: https://issues.apache.org/jira/browse/IGNITE-9026
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
We switched to CONTINUOUS mode based on the assumption that SHARED mode had
regressed in a way that allowed it to create many class loaders, and
eventually run out of Metaspace.
CONTINUOUS mode failed much sooner, and we were able to reproduce that
failure and identify bugs in the code. The
We are running in SHARED_MODE on 2.5, and are currently quite suspicious of
this change in 2.4, the essence of this change is, in SHARED_MODE , to just
skip the code that will "Find existing deployments that need to be checked
whether they should be reused for this request"
Is there a simple way inside Ignite to get a log of all SQL Queries against
the cluster, either in the debug logs or elsewhere ?This is not a easy
question to phase in a way that Google will find a useful answer.
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
The nested class hypothesis seems unlikely. We have 6000+
GridDeploymentClassLoaders on a node, because there are many instances of
"GridDeploymentPerVersionStore.SharedDeployment".
The userVersion is not changing, nor is the cluster topology.
I have enough data to debug this, just need some
Your original stack trace shows a call to your custom stream receiver which
appears to itself call invoke(). I can only guess that your code does, but
it appears to be making an call off node to something that is not returning.
2.4 should be OK.
What you showed that the stream receiver called invoke() and did not get an
answer, not a deadlock. Nothing looks particularly wrong there. When we
created this bug, it was our a stream receiver called invoke() and that in
turn did another invoke, which was the actual bug.
It
"When receiver is invoked for key K, it’s holding the lock for K." is not
correct, at least in the 2.4 code.
When a custom stream receiver is called, the data streamer thread has a
read-lock preventing termination, and there is a real-lock on the topology,
but DataStreamerUpdateJob.call() does
The documentation describes the use case where a node is stopped and removed
from the baseline, which reduces the number of backups/replicas when the
node is stopped.
I assume that there is no current code to support removing the node from the
baseline first, so that at least desired number of
It does appear that if the global concurrentLevel is not set, that it
defaults to # CPUs not 4 * # CPUs as documented here
https://apacheignite.readme.io/docs/memory-configuration#section-global-configuration-parameters
private long[] calculateFragmentSizes(int concLvl, long cacheSize,
Early on running on 2.3 we had hit a clear deadlock that I never root-caused,
where the cluster just stopped working. At the time I was use the same
DataStreamer from multiple threads and we tuned up the buffer size because
of that, and we were running against EBS, and perhaps with too short
When running on AWS, I found that what the "disk" that you are writing to is
the most critical issue for Ignite. EC2 instances with local SSDs have
about 20x the write rate as a multiple 3 TB GP2 problems, and using actual
disks (e.g., EBS) for Ignite Persistence storage is a non-starter.
When you set the stream receiver, an instance of its class is created and
serialized, which will also include any class it is nested in. On each
Data Streamer buffer, the serialized form of that class is sent. If the
class containing the stream receiver has a pointer that is not
We had done this to group all of the data that needs to be backed up onto the
SSD. Work also contains the log directory, and I haven't seen how to put
that elsewhere.
The introduction in 2.4 of Baselines seems quite helpful. If a node
restarts, it will avoid excessive rebalancing.
What is unclear from the documentation is what happens in the case where a
node fails, and doesn't come back. I'm assuming that in fact nothing
happens, except that the backups
Once a BinaryObjectSchema is created, it is not possible to change the type
of a field for a known field name.
My question is whether there is any way to determine the type of that field
in the Schema.
We are hitting a case were the way we get the data out of a different
database returns a
Just saw 2.4 release notes: Improved COUNT(*) performance
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
The server node was already active, and when I commented out
ignite.active(true) the client came up.
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
If I change userVersion in ignite.xml on the client to 5, when the docker
image is in SHARED mode, in order to ensure that our peer-class-loaded
classes are reloaded, I cannot start the client.
final Ignite ignite = Ignition.start(igniteConfig);
ignite.active(true); <<<
I fought with trying to get Ignite Persistence to work well on AWS GP2
volumes, and finally gave up, and moved to i3 instances, where the $ per
write IOP are much lower, and a i3.8xlarge gets 720,000 4K write IOPS vs on
the order of 10,000 for about the same cost.
--
Sent from:
I've started reproducing this issue with more statistics, but have not
reached the worst performance point yet, but somethings are starting to
become clearer:
The DataStreamer hashes the affinity key to partition, and then maps the
partition to a node, and fills a single buffer at a time for the
I've started reproducing this issue with more statistics, but have not
reached the worst performance point yet, but somethings are starting to
become clearer:
The DataStreamer hashes the affinity key to partition, and then maps the
partition to a node, and fills a single buffer at a time for the
I've done some additional testing. By shutting down another (the last)
client node that was running independent code, I was able to purge the bad
version of my code from the servers, while leaving the userVersion at "0".
Apparently in this case, the client nodes are "master" nodes. (The
I was trying to demonstrate changing classes on a client node so that classes
on servers get replaced with new code, but the symptoms make me believe that
I don't understand the rules at all.
The deployment mode is SHARED. I read the instructions and creates an
ignite.xml with a different
I made improvements to the statistics collection in the stream receiver, and
I'm finding an excessive number of retry's of the optimistic transactions we
are using. I will understand that and retry.
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
I got similar symptoms, but for a different root cause.
I was getting the original stack trace using the "cache" command in Visor,
but only when a client was connected. Using this command caused the clients
to disconnect. It turned out that I had enabled inbound TCP ports to the
servers, but
Opened a support tickets with GridGain and AWS. The former suggested this,
which helped:
20
21
22
23
24
25
The throttling
The only possibly different thing we are doing is using a VPC endpoint to
allow the nodes to access S3 directly, without having to supply credentials.
Is TcpDiscoveryS3IpFinder expected to work? I randomly get exceptions which
seem to be considered part of normal S3 operation, but are not
handled/retried.
com.amazonaws.services.s3.model.AmazonS3Exception: Slow Down (Service:
Amazon S3; Status Code: 503; Error Code: 503 Slow Down; Request
Yes. When I figured that out, my immediate reaction was to kill the running
ignite instance. But since that process was what was keeping the docker
container running, the container exits.
So instructions on how to run the benchmarks from inside the docker
container they are delivered in
As I understand more, I changed the subject.
The Docker image for Ignite 2.1.0 contains the ignite-yardstick benchmarks,
which include a ReadMe file that provides instructions that *do not work* if
you simply try to run them from inside the Docker Container where Ignite is
already running.
I
Sorry for the delay, spam filter.
I haven't been able to find an explanation for startup.sh, but when I login
to an EC2 instance created using the AWS community AMI for Ignite (which
only has docker and pulls ignite by the version you specify), I find
startup.sh in the current directory, and it
If I use this in Advanced Details
JVM_OPTS=-Xms1g -Xmx1g -server -XX:+AggressiveOpts -XX:MaxPermSize=256m
I get
startup.sh: line 15: export: `-Xmx1g': not a valid identifier
startup.sh: line 15: export: `-server': not a valid identifier
startup.sh: line 15: export: `-XX:+AggressiveOpts': not a
Now I see the "Type 'help "command name"' to see how to use this command."
below the list.
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
The documentation around CONFIG_INI says "The downloaded config file will be
saved to ./ignite-config.xml", except that did not happen. Therefore when I
started visor, I had no way to discover the cluster, so visor created a new
one. After much newbie confusion, I copied the config file
I'm also a newbie, but I'm running 2.1.0 and I seem to be hitting the same
problem, which sounds like it was fixed a long time ago.Is there
something else going on?
I've uploaded the 2 lines I pass when creating the EC2 instance from the AMI
as well as the config file I'm using, as well as the
65 matches
Mail list logo