nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Atom Pub Binding Super Slow

jradebaugh

Hi,

I was wondering if anyone had seen this or had any suggestions. I'm testing the CMIS war using Curl and also Apache Chemistry over the Atom Pub binding. I hit the entry point URL and immediately get a response. However, when I post the URL to return children for the root of the repository, it takes over 19 minutes to come back! It does eventually come back with a good response listing the handful of cabinets I have in my repo, but only after a very very long time.

$ curl -udmadmin:dmadmin http://localhost:8181/emc-cmis-ea/resources/repositories/repo1/objects/root/children/ > curlout.xml

% Total % Received % Xferd Average Speed Time Time Time Current

Dload Upload Total Spent Left Speed

100 223k 0 223k 0 0 197 0 --:--:-- 0:19:18 --:--:-- 60569

Find more posts tagged with

Comments

jradebaugh

I was able to get this to perform normally (time to execute went from 19 minutes to less than a second) using a filter like so:

$ curl -udmadmin:dmadmin http://localhost:8181/emc-cmis-ea/resources/repositories/repo1/objects/root/children?filter=cmis:objectId > curlout.xml

It still seemed to return a full set of attribute properties for each object so I'm not sure why exactly the performance changed so dramatically. I guess it's omitting the search for relationships or renditions or something that was hosing performance.

david.matheson

Are you sure it was the filter? Often when I run this test, it takes a loooong time, then I re-run it (still without a filter) and it is fast.

jradebaugh

You are correct. The filter thing was a red herring! Sometimes it's slow...seems likely that it's the first use...sometimes its fast, regardless of the filter.

samuel

Hi Jon,

If it is the first request for a service/resource that requires type information, it may take longer time because some schema information is loaded behind and it will be cached in server. For the following requests, the responding time should be much less than the first one.

david.matheson

Hi, Yue. Two issues:

I seem to get the problem after a while of inactivity. The content server and application server stay up in the interim, though. It's just the first request of the day, or a request after an hour or two of no requests. This happens regardless of platform. I make requests from browser and Objective C, same behavior. Jon tests using curl and sees this behavior.
Even the first request shouldn't take minutes, correct?

I'll try some SQL tracing today to find out where it's getting hung up.

Dave

jradebaugh

Just to confirming that I'm seeing the same results reported by Dave. The initial call after a period of inactivity is very slow. Anywhere from 2 to as long as almost 20 minutes as reported in my initial post.

samuel

Dave, for your two questions:

1) Correct. After one hour, the schema cache will be considered as stale and refreshed when necessary.

2) Bad news is that, in this early access release (EA4), this time may be up to several minutes, and good news is that we're working on this and it will be improved in our next release (in 6.7). In my test environment, this figure is down to around 40 seconds at present.

Additionally, besides this schema cache loading time, the entire AtomPub binding is expected to speed up a lot in the production release. Please stay tuned. Thanks.

david.matheson

Ok, so can we just run:

#!/bin/bash

while :; do

curl -I -udmadmin:dmadmin http://localhost:8080/emc-cmis-ea/resources/repositories/repo1/objects/root/children

echo "Waiting ten minutes..."

sleep 600

done

in the background all day? Will this prevent the schema cache from expiring? Will we be able to use the 6.7 release pointed at a 6.5 repository? Will that be supported?

Dave

JasonChen

Dave,

You can set property cmis.property_info.cache_expiration_after_x_seconds in cmis-runtime.properties to change expiration time for the schema cache. Please find details in CMIS Deployment Guide.

The new release will support 6.5 repository.

Jason

david.matheson

Running my bash script doesn't help, it still refreshes the cache every hour. Is one hour a reasonable amount for this timeout? What would be the impact of setting it to 24 hours instead of one?

Dave

JasonChen

david.matheson 编写:Running my bash script doesn't help, it still refreshes the cache every hour.  Is one hour a reasonable amount for this timeout?  What would be the impact of setting it to 24 hours instead of one?Dave

it still refreshes the cache every hour? Have you set the property?

If you set it to 24 hours, the cache will refresh every 24 hours.

Jason

david.matheson

No, I had not altered the property. I was letting others know that even with steady activity, this cache will refresh after a given amount of time (default one hour). There was earlier discussion that suggested it was after an hour of inactivity, and I was noting this is not the case.

My question was what potentially adverse impact it would have to set this value to 24 hours. If I set it to 24 hours and then add a property to a type, does this property not show up for 24 hours?

Dave

JasonChen

david.matheson 编写:  If I set it to 24 hours and then add a property to a type, does this property not show up for 24 hours?Dave

Yes.

david.matheson

Is there a way to force the cache refresh? What happens if I "SELECT new_property FROM custom_type WHERE stuff" before this refresh happens?

abc123

For the released CMIS, you are not able to force a cache refresh before cache expiration. Regarding the data dictionary update, either Content Server or DFC has its time-delay, you are not easy to find an ideal way to meet all your requests.

But the good news is we are adressing this issue for the next release.