PI to CONNECT release history

116010

ClearChangeTablesDays can now be set in the agent configuration file appsettings.json under PIAssetFrameworkServerNames. This new setting corresponds to the ClearChangeTables setting that can be configured in C:\Program Files\PIPC\AF\AFService.exe.config. By default, both ClearChangeTablesDays (in appsettings.json) and ClearChangeTables (in AFService.exe.config) are 7 days. If the user updates ClearChangeTables, users can now manually adjust ClearChangeTablesDays to be the same. Keeping them in synch does not change agent behavior other than messaging. For example, if the user changes ClearChangeTables to 1, then the user can also change ClearChangeTablesDays to 1 so that the agent will log a warning message in the event that the agent cannot retrieve change notifications for AF Elements for more than 1 day, indicating that some AF Element changes may be missing from the transfer.

116301

When performing data recovery (called Streaming Backfilling), the agent would transfer the same data multiple times, provided that the recovery period spanned multiple archives. For example, if the streaming backfill period had spanned 3 archives, the agent would send data for the first archive 3 times, the second archive 2 times, and so on. If the agent fell too many archives behind, it would never recover. This problem was more likely to occur when the agent encountered periodic disconnections from Data Archive combined with frequent archive shifts (1 per day) and high data rates (100K/sec). This problem has been fixed.

117894

Connectivity issues to Data Archive during the historical transfer step could result in data loss during the transfer. Customers who were affected by this problem noticed that the transfer progress appeared to be stuck on a particular archive. Eventually, the current activity of the agent transitioned to "Streaming only", indicating that the historical transfer step was complete, but the progress continued to indicate that the agent was still "processing archive x of y". When examining the actual data that was transferred to CONNECT, the customer might notice that most streams have all data transferred, as only a handful of streams would be affected over narrow time windows. This is because the same connectivity issues that caused the data loss also caused the built-in retry mechanisms of the agent to kick in shortly after the connectivity problems occurred. This problem has been fixed.

115912

The transfer could stop with an "Uncategorized Error" if error -11106 (Failed to lock archive-cache point) occurred while trying to get the archive file list Data Archive. This problem is fixed.

118022

Elements with templates would fail to be transferred when the description on an attribute was different than the description on its corresponding attribute template. This problem was due to a restriction in the assets service in CONNECT data services that required that the descriptions on assets and asset types must be the same, apart from the description on the resource itself. For example, the asset would be successfully created if merely the asset description was different than the description on the asset type, but if an asset metadata item had a description that differed from the corresponding description on its asset type, the asset would fail to be created. This same limitation applied to stream references and their corresponding type references. This problem is now fixed in CONNECT data services.

PI to CONNECT Agent Installation

2.4.2520

71702

The PI to CONNECT Agent now supports legacy Data Archive server IDs originating from Data Archive version 3.4.370 or earlier. Previously, the agent would exit when configuring a transfer against a Data Archive with a unique ID in the old format. The old format is now converted into a GUID in the same manner that the AFSDK converts a legacy ID into a GUID.

111198

The agent would terminate unexpectedly when PI Update Manager is unresponsive during data recovery operations. For example, if the agent receives a data loss notification from PI Update Manager, the agent will restart the transfer to backfill any data that may have been lost. When stopping the transfer, if the agent received a "[-10722] PINET: Timeout on PI RPC or System Call" from the Update Manager when removing a signup, that error is not properly handled and would result in the agent terminating. This problem has been fixed.

112269

An active PI to CONNECT transfer would stop, requiring manual intervention to restart the transfer, if error -10007 "Bad or Null Pointer" was returned when getting an archive list from the Data Archive server. This error can occur at any stage of the transfer (whether the historical step is active or not) when archives are being unregistered on Data Archive. This problem is likely to occur when archive files are unregistered on a regular basis. This problem has been fixed.

113002

After the PI to CONNECT Configuration Utility connects to CONNECT data services, the user is presented with a list of namespaces in a dropdown. Prior to this fix, the entries in the list would correspond to the namespace description instead of the namespace name. For most users this would not cause confusion because the name and description are the same by default. However, if a user renames folders in CONNECT, only the namespace name gets renamed. In this release, the namespace name is displayed instead of the description.  

113303

After a silent installation of the PI to CONNECT Agent, the configuration utility would exit unexpectedly the next time it was launched. This was because the silent installation did not include an AlternateDisplayName in the appsetting.json for the Data Archive or AF server. This problem has been fixed.

113925

The transfer would stop unexpectedly when no archives are registered corresponding to the time range of the historical transfer step. The probability of this happening would increase significantly when the historical transfer step was complete and when a customer was doing archive maintenance, such as unregistering and moving archives to a separate disk. Event ID 90 would appear in the PI to CONNECT Event Viewer log, with the following message text:
System.ArgumentOutOfRangeException: Index was out of range. Must be non-negative and less than the size of the collection. (Parameter 'index')

This problem has been fixed.

114506

When CONNECT data services was unresponsive for over 10 minutes and when streaming updates had filled an internal agent queue, the transfer would stop, and an uncategorized error would be reported in the cloud, requiring manual intervention to restart the transfer. This problem has been fixed.  

PI to CONNECT Agent Installation

2.3.2349

111668

When transferring AF element templates to CONNECT data services both dynamic and static AF attributes are now included in the transfer, allowing both type references and metadata, respectively, to be included in the resultant asset type. Previously only static AF attributes were included in the transfer, which resulted in asset types that included only static metadata values.

111398

Support was added to the PI to Data Hub Agent service to enable asset/asset type PATCH operations. This enhancement allows a user to edit asset and asset types in CONNECT data services without losing their edits when the corresponding AF elements or AF element templates are re-transferred because of on-premises changes to the AF database. This means that on-premises changes will continue to be synchronized while also preserving additional independent changes that are applied in CONNECT data services.

105315

The streaming step can stop due to an unexpected exception. These unexpected exceptions cause the transfer to stop and will not restart automatically. In this release, additional log messages were added to track down the source of these errors.

110973

If a user enabled verbose logging for PI points or AF elements in the CONNECT data services portal, the agent would continue logging verbose messages for these points/elements even after disabling verbose logging in the portal, until the original expiration date that was set for verbose logging expires. This problem has been fixed. Disabling verbose logging in the portal now causes verbose logging to stop immediately.

PI to Data Hub Agent Installation

2.2.2174

100961

In some cases, the license file for a PI Data Infrastructure – aggregate tag unlimited server included 16 additional PI tags. The agent update accounts for that difference and will now consider a license with MaxPointCount greater than or equal to 999,999,999 tags as unlimited and report the actual tags created, not the MaxPointCount as tags used.

PI to Data Hub Agent Installation

2.2.1567

97181

The default operation timeout for the agent has been increased from 30 seconds to 300 seconds (5 minutes). The default was changed because Data Archive RPCs typically have timeouts of 270 seconds (4 minutes and 30 seconds). The mismatch in timeouts lead to cascading problems and data loss. The new default takes effect only for new installations and must be manually applied for upgrades. Customers upgrading to this version should consider changing OperationTimeout to 300 in C:\ProgramData\OSIsoft\PIToOCS\appsettings.json to avoid putting unnecessary load on Data Archive when timeouts occur.

97168

When sending historical data to the cloud, messages are now batched together. This significantly improves performance under some conditions. The biggest improvements are seen when there is high latency between agent and cloud (such as a transfer from an agent in eastern US to a namespace in western US) and when a large number of string or digital PI Points are being transferred.

97704

Added a log message that reports the breakdown of the PI points included in the transfer by point type.

97705

Added a new appsetting.json setting, HistoricalReadTaskCount, to allow the preferred number of Data Archive read threads to be configured. Typically, this value should not be changed except when advised by technical support.

97706

Changed the behavior of an existing appsetting.json setting, HistoricalSendTaskCount, so that it can be increased to at least 12, regardless of processor count. This setting can be used to increase the parallelization of sends to the cloud, possibly improving performance in high-latency scenarios, such as transferring data from an agent in eastern US to a namespace in western US. The benefits of increasing HistoricalSendTaskCount should be weighed against the increased likelihood of sending data out of order to the cloud. Typically, this value should not be changed, except when advised by technical support.

97195

Added extra logging around the "transport message too large" errors so that root cause can be determined from the log message. Example error: Transport message too large to persist in Event Hub and will not be sent! Actual: 253075 Max: '235929'.

97857

Treat license count of 999,999,999 as an unlimited license for purposes of charging Flex credits for AVEVA PI Data Infrastructure – aggregate tag PI Servers.

97183

The agent now caps the maximum events per query to 150,000 events when timeouts occur or errors occur that might indicate that Data Archive is busy. Previously, capping the events to 150,000 was done only when ArcMaxCollect was exceeded (error-11091). The new behavior is to cap requests at 150,000 events when the following occur:

• PI timeout exception

• Error -11140 is (Archive_MaxQueryExecutionSec exceeded)

• Error -10746 (MaxMessageLength exceeded)

  A customer ran into a situation, where switching to 150,000 event chunks would have reduced pressure on Data Archive. They were doing a backup of Data Archive and they were querying streams with over 1 million events for a particular archive. They were running into timeouts on the agent side (PI timeout exception) and also exceeding Archive_MaxQueryExecutionSec on the server side.

74686

The transfer could appear to get "stuck" during the Historical transfer step. For example, if there were 10 archives in the transfer, the progress reported in the CONNECT data services portal could indicate, indefinitely, that archive 9 of 10 was being transferred. Although this was only an issue with the progress report, this problem, combined with item 99974 below, would make it appear that the agent was still attempting to fill data gaps when, in actuality, the historical portion of the transfer was already complete.

99974

During the historical transfer step, data gaps could occur if there were communication problems with Data Archive that lasted longer than 5 minutes. This problem has been fixed. The agent now retries indefinitely when retriable errors occur.

78720

If Data Archive was shut down during agent registration, the agent would stop unexpectedly. This problem has been fixed.

98497

When messages with event IDs of 43 and 122 appeared in the event log, the agent could get into a bad state and consume all memory on the computer. The memory growth problem has been mitigated. The agent will now log an error message and move on to the next message in the queue.

95557

Extra values or incorrect values can be sent by the PI to Data Hub Agent after a range delete operation followed by a backfill. This operation is common in AF Analytics when recalculating analyses. The incorrect value problem has been fixed in the agent. Note, however, that there will still be some cases in which extra values appear in the cloud, though these extra values within the compression deviation tolerance. This work item only addresses agent side fixes for range delete. There was a separate bug fix on the cloud side, as described by work item 2824330. Specifically, the cloud service was not processing range delete or individual delete events in the order sent by the agent. Both of these problems are now fixed.

98282

When starting a transfer, a transfer job can fail to start if the transfer job status is requested from the cloud service before all steps are added to the transfer. This bug could affect the startup phase of any transfer. This problem has been fixed.

97194

Fixed thread safety issue when reporting agent progress. The error was reported in the log was: "System.InvalidOperationException: Collection was modified; enumeration operation may not execute."

97167

Transfer edits could cause a very large number of time ranges to be resent to CONNECT data services. This problem has been fixed.

96143

Windows Authentication is now favored over PI trusts for connections to Data Archive. The authentication order was reversed in all previous versions (2.2.1163 and earlier), where the agent would first try to connect using a PI trust (less secure) before trying to connect using Windows security (more secure).

PI to Data Hub Agent Installation

2.2.1539

PI to Data Hub Agent Installation

2.2.1163

PI to Data Hub Agent Installation

2.1