Replication nodes

Similar Questions

• ISE Admin error replication node

  Hi all

  Sometimes I receive this alarm:

  Alarm

  Has taken place at:	Thu Mar 20 09:20:10 BRT 2013
  Cause:	Replication stopped
  Details:	Order of replication for the host (node Secundary Admin) PANVMGP3301B

  Today, I go to Administration-> system-> deployment and I can see my secondary node of Admin/analysis with status "DISABLED REPLICATION" picture to attach it.

  I can force the synchronization for the primary and secondary Admin nodes? How can I fix?

  TKS!

  Yes, something is preventing your nodes to stay in harmony and accordingly, the nodes stopped trying to syncup. You will need manually synchronize the nodes. Go to Administration > deployment. Select / check all nodes, then click on the button "Syncup" above the personas.

  Thanks for the note!

• Scalability of replication

  Hello
  
  Congratulations on the new feature of replication, it looks really good!
  
  How scalable is replication? What is an architecture target for her?
  I understand that everything depends on the application - the amount of data, how many Scriptures/readings etc but I am interested in some tips gross (I think you have tested different scenarios).
  After reading the documentation, it seems to me that it will not probably evolve more, so let's say only 10 knots (because the master must propagate changes to all replicas)?
  
  Best regards
  Jarek

  Jarek,

  There is no specific limit on the number of nodes. The only limitation is one of the resources such as bandwidth or CPU network master node, so the question is, as you say, how a replication group uses these resources.

  We look at scalability as a two part question - write scalability and read scalability. For this first release, we are focused mainly on the configurations of a dozen knots and under. With the unique master architecture, write throughput in a single replication group is limited by the capabilities of the master node. Scaling above which is possible to partition the application data, and who currently is something that must be implemented in the application itself.

  Read operations that require a certain level of consistency (which that data are up to date) are, as you request, limited by the fact that the single master propagates changes to all replicas. With this in mind, we put in preparation for future improvements to stoke the replication stream and leave the replica nodes to propagate the changes to other nodes of the replica, but we will wait that improving the feedback on the current function to determine what priority would have. In general, replication node readings have no additional charge above a non-HA configuration.

  I'm not quite sure what you mean by a target architecture. Perhaps mean you an application architecture, or maybe the platforms we tested on? As for the architecture of the application, we thought that on the cases where the application focuses mainly on the availability and trafficking replicas as hot Standby, as well on cases where read scalability was desired, and the aftershocks are actively maintenance requests. The two scenarios have been targeted.

  With respect to the platforms, we tested with a range of materials, ranging from multi core to dog ears computers portable server class machines and also a variety of configurations of the AWS. Measures of importance so on the application and the platform, it would be probably more instructive to briefly describe what our measures were:

• write in a replication of group performance must meet or exceed a non replicated environment performance, synchronously sustainable I alone, and on most platforms, we met this goal. As you guess, it will degrade as the number of aftershocks increases to the point where dominate latency General fresh network and receipt.
  
• performance in reading on a replica should be comparable to that available on a non replicated environment I with a load of single-threaded update similar to the update of the master stream.

  We found that many factors, ranging from hardware, platform configuration, size of replication group has had an impact. In the case that you are talking about, a group of most important replication often has better write throughput that a group of 2 to 3 knots, even if the master node must be propagated to multiple replicas, because a larger group may be able to more effectively acknowledge validations. An intensive application of writing very saturate the master more quickly, while an application-oriented read more may be able to use other nodes. It is best to set all, first a request for best performance in a non replicated system and then enable replication, so it is easier to understand the impact.

We are really looking forward to comments on what people can do with the system, and what are the types of improvements would be useful.

Hope that helps,

Linda

• Cleaning store policy

  Hello

  We have developed an application that writes binary data to a store of NoSql as LOB values (about 300 KB each).

  The data is first stored and read by another thread and finally deleted from the store.

  We see that the size of the store on the file system keeps growing and unused space is never released. In the course of a long ordeal, we have a complete file system!

  What is the policy of the cleanest threads? How can configure us the parameters of policy?

  The store is configured with replication factor = 3

  We tried to set the parameter "serviceLogFileLimit" but nothing has changed and storage has continued to grow over the configured limit...

  Thank you

  Giovanni

  We have developed an application that writes binary data to a store of NoSql as LOB values (about 300 KB each).

  The data is first stored and read by another thread and finally deleted from the store.

  We see that the size of the store on the file system keeps growing and unused space is never released. During a long test, we got a full file system!

  What is the policy of the cleanest threads? How can configure us the parameters of policy?

  The store is configured with replication factor = 3

  We tried to set the parameter "serviceLogFileLimit" but nothing has changed and storage has continued to grow over the configured limit...

  There are a least two uses unrelated to the word "log" in the system, and I think that serviceLogFileLimit is not related to one that's caused you problems here.

  The serviceLogFileLimit setting refers to store debug logging output files, that contain log entries with textual information on the events in the system, as the warning events that you mention.  These files are located in the subdirectory of the directory top level store log and have names like admin1_0.log or sn3_0.log.  This setting limits the maximum size of each of these files, which controls when the logging system switches to another log file.

  The log files, you are having issues with that are the files of newspapers I, what are the files that I use to store the changes made to the database.

  If you were really more space drive due to the recording of debug files, you probably would also change the 'serviceLogFileCount' parameter, which controls the number of copies of a given debug log file is maintained.  The default values of these parameters are serviceLogFileCount = 20 serviceLogFileLimit = 2 MB, for a total of 40 MB per set of log files.  There is a set of files by RN, admin and SN, as well as newspapers combined for the entire store.  To 40 MB for each of these sets, however, who only use it for storage.

  To add more details... log store is filled with warnings like this:

  2015-02-02 13:51:55.912 UTC WARNING [rg2 - rn3] I: vacuum cleaner 5 files not deleted because they are protected by replication.

  2015-02-02 13:51:55.925 UTC WARNING [rg2 - rn2] I: replication prevents deletion of 5 files by vacuum. Start the file = wedges to 0 x 0 CBVLSN 1, end file = 0 x 8 holds last VLSN 6 712

  As you say, these warnings mean that the files of newspapers I are selected for a reason any.

  The release of ping you provided confirms that replicas are caught up with their masters, because "sequence number" values are the same for all members of each burst.  If this isn't what causes the files to keep.

  We run another test which consisted of the following repeated looping: store a LOB, read it, delete it.

  Given the size of the LOB (300KB), this operating model will certainly generate a lot of log file data I.

  My guess is that the problem is related to the behavior of I which retains a minimum number fixed VLSNs.  There is an IC undocumented parameter that specifies the minimum number of VLSNs, which is by default 20000.

  With objects of 300 KB and 20000 VLSNs corresponding to about 10000 objects, i.e. about 3 GB of space dedicated to these VLSNs selected.

  I think it would be interesting to experiment with the reduction of the value of this setting to see if this results in the deleted log files.

  The name of the parameter I is:

  je.rep.minRetainedVLSNs

  It would be wise to try to reduce the value of its default value of 20000, say, 2000, which should decrease the minimum space held about 300 MB.

  The Administrator's guide provides information about the store settings:

  http://docs.Oracle.com/CD/NoSQL/HTML/AdminGuide/SetStoreParams.html

  You can set the value of a parameter I by changing the value of the replication node setting configProperties.

  You can check the default value of the parameter configProperties starting up of the CLI admin and doing:

  view the policy settings

  To check the value on a given RN, say rg1 - rn1, if the different nodes have different parameters, do:

  display settings-service rg1 - rn1

  Add the new parameter to the value of the existing setting.  For example, if you see:

  configProperties = foo 30; bar 22;

  You can use:

  configProperties = foo 30; bar 22; je.rep.minRetainedVLSNs 2000;

  Set the current value for all AI by:

  plan change-settings-all-rns-wait configProperties - params = "foo 30; bar 22; je.rep.minRetainedVLSNs 2000; »

  You must also change the default value by:

  change-policy - params configProperties = "foo 30; bar 22; je.rep.minRetainedVLSNs 2000; »

  If please give that a try and let us know if this solves the problem.

  -Tim

• How YCSB with Oracle?

  Hi all

  I followed the instructions available here: Installation Configuration to install Oracle NoSQL and can even run the HelloBigDataWorld smoke test.

  However, when I try to run the YCSB, they would give me an exception "could not contact any RepNode to: [server0:5000]".

  Some background information:

  I have access to 4 servers, and I'm supposed to configure Oracle NoSQL on 3 of them (server0 ~ 2), then run YCSB on the last one (3 Server).

  Here is my result to "see the topology:

  kV-> see the topology

  Ranger = mystore numPartitions = 30 sequence = 40

  DC = [dc1] name = name repFactor = 1

  SN = [sn1] dc = dc1 server0:5000 capacity = 1 FOOT race

  [rg1 - rn1] RUNNING

  No info available performance

  SN = [sn2] dc = dc1 server1:5000 capacity = 1 FOOT race

  [rg2 - rn1] RUNNING

  No info available performance

  SN = [sn3] dc = dc1 server2:5000 capacity = 1 FOOT race

  [rg3 - rn1] RUNNING

  Avg latency = 1,9051613 ms avg latency multi-op single-op = 0.0 ms

  Shard = [rg1] partitions num = 10

  [rg1 - rn1] sn = sn1

  Shard = [rg2] partitions num = 10

  [rg2 - rn1] sn = sn2

  Shard = [rg3] partitions num = 10

  [rg3 - rn1] sn = sn3

  and here's my YCSB config file:

  # Name of the store

  storeName = mystore

  # List separated by commas of the host/port to support pairs

  helperHost = server0:5000

  I have no problem to run the server3 HelloBigDataWorld test:

  Hello $KVHOME/lib/kvclient.jar:$KVHOME/examples me@Server3:~$ java - cp. HelloBigDataWorld-host server0-port 5000 - shop mystore

  Hello world data!

  But with the YCSB:

  me@Server3:~$ $YCSB/bin/ycsb charge Pei Pei loads/workloada nosqldb.properties nosqldb

  Commencement of the trial.

  com.yahoo.ycsb.DBException: oracle.kv.FaultException: 11gR2.1.2.123 NoSQL DB, Oracle could not contact any RepNode to: [server0:5000]

  No idea why? I tried the latest version of Oracle NoSQL and the oldest version that I could find, and both gave me the same problem...

  Moreover, to make this point, should I configure 3 data centers with 1 storage node each, or 1 data center with 3 storage nodes (I have no need of a replication node - it's only for benchmarking).

  Thank you!

  [T] hen I try to run the YCSB, they would give me an exception "could not contact any RepNode to: [server0:5000]»

  [...]

  me@Server3:~$ $YCSB/bin/ycsb charge Pei Pei loads/workloada nosqldb.properties nosqldb

  Commencement of the trial.

  com.yahoo.ycsb.DBException: oracle.kv.FaultException: 11gR2.1.2.123 NoSQL DB, Oracle could not contact any RepNode to: [server0:5000]

  No idea why? I tried the latest version of Oracle NoSQL and the oldest version that I could find, and both gave me the same problem...

  I was able to reproduce this problem, and the issue is an incompatibility in the client introduced between versions 1 and 2.  The predefined Oracle NoSQL provided with YCSB 0.1.4 database driver is for the version 1.2.123, but the first version of Oracle database currently available NoSQL is 2.0.39 and server of this version is incompatible with the version 1.2.123 client used by the YCSB driver.

  I was able to get things by updating them my YCSB 0.1.4 source to designate the 3.0.5 client referenced here:

  NoSQL Oracle on Maven https://community.oracle.com/thread/3575499 >

  I had to change the top-level pom to comment on a bunch of other modules (to avoid other problems) and activate the nosqldb one.

  I changed the nosqldb/pom.xml to have:

  com Oracle.kV

  kvclient_3.0.5

  3.0.5

  And:

  Central

  http://repo1.Maven.org/Maven2

  I did a 'mvn package ".  When I used this rebuilt version, the YCSB run succeeded.

  Moreover, to make this point, should I configure 3 data centers with 1 storage node each, or 1 data center with 3 storage nodes (I have no need of a replication node - it's only for benchmarking).

  Separate data centers using only affects the way that replication nodes are distributed to survive data center failures, which shouldn't matter to a performance index.  However, the replication factor - the number of nodes of replication by shine - will impact performance.  The master node handles all update requests, but replicas can handle queries from reading, which takes the load off of the master and should greatly increase performance.

  -Tim

• RGST or SYLVAIN

  Hi Charles,
  
  I can see RGST (Table to gratify the replication group) in http://www.oracle.com/technetwork/database/nosqldb/learnmore/nosql-database-498041.pdf page 10 ("Pilot Client" section for example).
  I see SYLVAIN (State of the replication node Table) in http://hpts.ws/sessions/cwl-hpts-for-website.pdf page 18 (see paragraph "data location: Rep State Table node").
  RGST and SYLVAIN the same thing?
  If this is not the case, what is the difference? Where is stored SYLVAIN? At the level of the customer?
  Thank you

  Yes, the RGST and SYLVAIN are the same thing.

  Charles Lamb

• White Book of Oracle database NoSQL (application)

  Hi all
  
  Please, I would like to have an answer to the following questions:
  (1) what do you mean by topology - when you say "the client driver maintains a copy of the topology?
  (2) not working in the NoSQL of Oracle database as a client/server? Do we not have to install the client on every PC that needs to use the store? How will we if we have a new PC? End users have a customer on their PC? I don't really understand where the client has to be installed.
  (3) you say: '30 replication groups are stored on nodes of storage 30, distributed on the two data centers' why 30 storage nodes? It should be 90 knots since we have a factor of replication of 3.
  (4) how can the customer get the answer to a read request? The customer directly reads the response of a log file on one of the replication node? Those that read the request was transmitted?
  Thank you

  user962305 wrote:
  Hi all

  Please, I would like to have an answer to the following questions:
  (1) what do you mean by topology - when you say "the client driver maintains a copy of the topology?

  Topology is the location of the nodes in the system and that the partitions that they contain. By knowing the topology, the customer can contact the node holding the requested data directly.

  (2) not working in the NoSQL of Oracle database as a client/server? Do we not have to install the client on every PC that needs to use the store? How will we if we have a new PC? End users have a customer on their PC? I don't really understand where the client has to be installed.

  Any application that wants to use the NoSQL database must have the kvclient - M.N.P.jar in the classpath. This pot located in the KVHOME/lib directory.

  (3) you say: '30 replication groups are stored on nodes of storage 30, distributed on the two data centers' why 30 storage nodes? It should be 90 knots since we have a factor of replication of 3.

  Can you tell me the location of the book and what page you are talking about?

  (4) how can the customer get the answer to a read request? The customer directly reads the response of a log file on one of the replication node? Those that read the request was transmitted?

  If a request has been sent, the response is returned to the client via the transfer node.

  Charles Lamb

• ReplicatedEnvironment: Master asks of preference and re-election.

  Hi all
  
  I have two questions about replicated environments.
  
  Assume that a group of replication with 3 nodes likely to be elected.
  Lets name them "Higher", "Medium" and "Low" as a way to describe his "power" in terms of better CPU, more memory, etc.
  
  1 ^ st ^ question: is there a way to specify an order or preference when choosing a master in an election on a ReplicatedEnvironment?
  
  Now suppose that after a run time where 'Superior' was the captain, it breaks down due to a problem and a new election is performed.
  After a moment, the problem is solved and 'Superior' is running as a replica.
  
  2 ^ nd ^ question: is it possible to force thus higher re-election could again be the master?
  
  ARO
  / Caesar.

  Caesar,

  A simple way to designate a given node, as the captain when the group is first created should simply bring this node first. Since this is the first member of the group, it is by default, the master.

  Once the Group was filled with nodes, teachers are chosen by election. The elections were won by the replication node that has the latest data. Tie, an arbitrary but predictable attribute is used to break the tie. In the current version, this switch is an artifact of the implementation, and it is not available through the public API.

  We thought to expose the notion of priority of the node in the public API, to be used as a factor in the breaking of the tie. Another option is to make priority over a tiebreaker and let the application dictate the master. This last option is much more problematic than the first, because it can cause data loss, if a node is late chosen as master. In your case, the first option would serve the purpose you want?

  Note that a group of ongoing replication of execution with nodes of different resources can curtain at the performance of the group as a whole. Nodes are reproducing all the operations in the master, so the flow of the group will be blocked by the slower node. If the application is to pump a volume of operations which could be resolved by a slow reply, it may also cause com.sleepycat.je.InsufficientAckExceptions thrown for postings of transactions.

  Practically speaking, as a slower knot is likely to have less common data, it is less likely to be chosen as master.

  Another, more great feature on our list is the possibility to designate a node as not likely to be elected. The idea here is to support a replication group which consists of N nodes likely to be elected with similar resources, as well as a set of nodes of M that are perhaps less able, or remote. As always, we cannot talk about specific release dates.

  We would welcome some comments.

  Kind regards

  Linda

• Replication Failover clustering & Hyper-v on a two-node cluster

  Hello

  We have two identical servers.  One to use as a main server, which houses two virtual machines.  The other should be used as a back up in case the main server is not somehow.  We planned on using a solution of third-party software to back up our virtual machines and launch them in the event of failure of the principal server.  However, we discovered just gave the failover clustering in Windows Server R2 2012.

  We have tried to set up a cluster 2 nodes, with our virtual machines properly replicated to the backup server.  Excited by the present, we tried then to simulate failure of the primary server (by pulling on the cable network).  We were a little disappointed to find that the backup server is not automatically run virtual machines.  We did a survey and read than a solution of two nodes requires storage space for additional network (in addition to our two servers).

  I wonder if that's okay?  In other words, a two-node failover cluster requires separate network storage space.  If this is not correct, can someone point me to a set of instructions to correctly configure a two-node failover cluster that will automatically launch the virtual machines on the server backup (in the case of a failure of the principal server).

  Thank you

  Mike Goldweber

  Hello

  Post your question in the TechNet Server Forums, as your question kindly is beyond the scope of these Forums.

  http://social.technet.Microsoft.com/forums/WindowsServer/en-us/home?category=WindowsServer

  See you soon.

• Dell FS7610 replication question

  We currently have a FS7610 NAS and an EqualLogic PS6510E (with ~ 100 TB useful).  Our company just bought another FS7610 and PS6510 for replication purposes (the two new units will reside in another data center) and in the coming months we'll probably buy a bunch more (FS7610 and PS6510s) which will also replicate each other.  At this point, I don't know what to expect, but I am a little concerned about bandwidth Layer 3.  I talked to a couple of sellers of Dell and got two different answers on this... but on a FS7610, replication happens on the front-end or back-end?

  Thank you in advance!

  In the case of the 76xx, replication is performed by the FS, not EQL table nodes, and it uses the subnet of the iSCSI network to perform replication.

  Kind regards

• Comments of the NAC Server replication

  I have reimagee one of our NAC comments 2 servers and I now want to restart replication.

  To set up replication using two comments Cisco NAC servers, the IP address of the remote server is configured on each server, and one of the following 2 options must be selected:

  "This node contains the data.

  "This node will copy the data of the other node.

  Is 'this node' the remote server or local server (where the configuration is done)?

  The initial replication is configured by putting one of the Cisco NAC comments servers to copy all the data from the other server of comments. The comment server that is configured to copy the data from the other device is first set to remove all its own data. This ensures that no conflict exists. Cisco recommends the introduction of replication during the initial installation of the server Cisco NAC comments, or when you add a new comment server to an existing implementation. Once one of host servers received a copy of the data from the other device, they are synchronized and replication is enabled. All data that are updated on a single server of comments are then automatically replicated to the other server of comments.

• ISE 1.2 CWA with several Ssnp - SessionID replication / Session expired

  Hi all.

  I have a nodes of Services (2) policy (NSP) in a deployment of ISE launched 1.2 patch 1. We use Wireless MAB and CWA on Wireless LAN controllers v3.3.3 running 5760.

  We hit a problem in which a client comes first MAB and then is redirected to a custom portal CWA. The customer then receives a message from the Session has expired. This seems to be related to the fact that the CWA is technically an authentication of 2 floors (MAB by the WLC) and then of the CWA by the customer. Specifically, it seems to happen when the WLC makes his request to access PSN - 1 RADIUS MAB and then the customer comes to PSN - 2 to finish the CWA. This problem does not occur when a NHP is being used and all traffic authentication (RADIUS MAB and CWA) caters to a unique PSN.

  Customers solve the COMPLETE domain name in the redirect URL using DNS public and a public DNS zone file (let's call it cwa - portal example .com). CWA - portal example .com has two records for the two nodes of NHP. DNS responds to queries using Robin DNS.

  I have the Ssnp configured in a node group for replication of session information between PSN, but this does not seem to make a difference in the behavior.

  So I ask:

  What is the architecture recommended for CWA when you use more than one PSN? It seems that you must keep the stream two authentication pinned together so that they both hit the PSN even when you use more than one PSN in a deployment. A load balanced on the SessionID string balancer comes to mind (demand of RADIUS MAB and contain both the CWA URL this unique by client SessionID), but that seems awfully oversized for a seemingly simple problem. On the other hand, it seems also that by using a configuration node group should easily be able to reproduce customer SessionID to all nodes in the deployment, so that is not a problem. That is, if the WLC authenticates MAB on PSN-1, then PSN-1 should talk the group node such that when the client CWA on PSN - 2, 2 - PSN responds with a Session expiration message.

  Is there a Cisco documentation which talks about this?

  Maybe in relationship:
  https://supportforums.Cisco.com/discussion/12131531/ISE-12-guest-access-...

  Justin

  Hi Justin,

  Node groups are mainly used for redundancy of the sessions that are waiting for status.  Thus, because the controller is configured to use the PSN-1 as the first RADIUS server, PSN-1 will be the session on the client information.  This information is not shared with PSN-2 that's why you see "expired session".  In short, the node that processes applications MAB, must be the node that serves as the personalized Portal.

  Round robin DNS is preferable for use with the sponsor of the portal and portal of my devices with an FQDN that is similar to sponosr.example.com and mydevices.example.com.  For CWA, a load balancer is the best option if you want to use multiple Ssnp.  Aaron Woland wrote and article covering the ISE and the load balancing.  F5 has also some useful information on how to configure their loadbalancers with Cisco ISE.

  Kind regards

  Tim

• Cannot save an ACS secondary for replication of ACS primary 5.2.

  Hello

  I hope someone can help me.  Currently, I have two devices Cisco ACS and both are classified in the PRIMARY.  The first ACS is running version 5.2.0.26 while the second ACS is running version 5.3.0.40.

  My original thought was to install the first ACS and do serve primary and have it replicate its data on the ACS SECONDARY.  Somehow, after installation, the ACS are now listed as PRIMARY.  When I go into secondary ACS under Deployment Options to try to save it in elementary school, I get the following error message:

  "This failure has occurred.  Failed to authenticate with node.  Your changes have not been saved. »

  Even if I try this GBA primary to save it for the secondary ACS, I get the same error message.  I tried all passwords including the credentials of the admin super user, my credentials for the administrator and the credentials provided to SSH in ' GBA and nothing is helping.

  Reading online, I read there was a way to remove an ACS secondary, but I don't have the ability to add this server in the primary for "bump it down" to a secondary antibody hoping to save it for the primary ACS.

  If anyone can give me some pointers, I would greatly appreciate.

  Thank you, and all have a wonderful day.

  THERE

  Yvonne,

  If the identifier is the same then definitely replication does not work, you will not be able to enroll in primary school if the license is the same. The good side is that you have the other license, you only need to install.

  However I have more bad news, the only way to re - install a license file in ACS 5.x uses the CLI command 'acs reset-config', but it will also delete all of the configuration that you have on this server, except the network configuration (IP, gateway, DNS, etc.)

  After entering this command if you are trying to access the GUI, you should not use the name of user and password acsadmin/default, then you will be asked to locate the license file.

  Here is a document with this information where you need it:

  http://www.Cisco.com/en/us/partner/docs/net_mgmt/cisco_secure_access_control_system/5.3/user/guide/my_wkspc.html#wp1052906

• ISE 1.1.3.124 secondary node not accessible after registration

  G ' Day all,

  I am constantly given the State of synchronization and replication for my secondary admin/Monitor node in the head node as a node unreachable. The secondary has always think it is offline. When I run the tool diag ISE connectivity tests, that I can ping successfully the devices together using both hostname and ip and the nslookup works also very well between the two nodes. Ping and nslookups also work different networks within the environment. The two nodes are in the same vlan on a pair of vss 6500, but on different switches of the pair. I am new to ISE, so any help is greatly appreciated.

  Thank you all.

  JS

  Sent by Cisco Support technique iPhone App

  Hi James,

  Node added to distrubited deployment but Sync never ends.

  Normally, this means that the TCP 1521 port is blocked between the nodes or the internal database user password was not set to the same value between the primary and the secondary. (I know you have already completed this step)  The user database can be reset on the CLI with the following command: reset-passwd ise internal-database - application user.

  Change the internal database user password restart the ISE services.

  Additionally, make sure that the secondary node is accessible on the following ports in both directions:

  -TCP 443 (SSL/SOAP)

  TCP - 1521 (database listener)

  -ICMP (Heartbeat)

  Jatin kone
  -Does the rate of useful messages-

• ISE PSN node will not be joining the cluster

  Hi all

  Has anyone seen a problem where an NHP cannot join the cluster?

  We join node of PSN

  -Node is saved successfully (current synchronization)

  -1 hour later - node replication failure.

  -Replication synchronization failed because the secondary database is down

  I have a client where admin node and PSN are separated by the firewall.

  We let in two directions

  Admin <-->PSN

  ICMP

  HTTPS

  1521

  Firewall not showing drops.

  DNS and NTP are ok.

  Current topology is 1 NHP, 1 Admin node.

  Works very well in our test lab, but not clients environmnet.

  See you soon

  Peter.

  Thank you for the update we and good work on the search for the solution! You should probably mark it as resolved now

  In addition, it is quite rare (at least for me) for nodes of ISE to be separated by firewalls. There are a lot of ports/protocols that must be opened between them is usually more of a pain to manage. In addition, sometimes ports will change too. For example, the fueling port agent has been changed not too long ago...

  Thanks for the note!